diff --git a/bench/results/20261002-004441-jebadiah-4b-v2-spark-4-baseline-beside-nemotron-nano-4b-decide.json b/bench/results/20261002-004441-jebadiah-4b-v2-spark-4-baseline-beside-nemotron-nano-4b-decide.json new file mode 100644 index 00000000..cbaf0492 --- /dev/null +++ b/bench/results/20261002-004441-jebadiah-4b-v2-spark-4-baseline-beside-nemotron-nano-4b-decide.json @@ -0,0 +1,2670 @@ +{ + "schema": 1, + "stamp": "20261002-004441", + "label": "spark-4 baseline beside nemotron nano 4b", + "model": { + "id": "frontier-infra/jebadiah-4b-v2" + }, + "placement": { + "engine": "vllm", + "node": "Spark-4-GX10", + "gpu": "NVIDIA GB10", + "ainode": "0.5.33", + "tp": 1, + "gpus": 1, + "stacked_with": [ + "frontier-infra/jebadiah-9b-v2" + ], + "port": 8003 + }, + "settings": { + "backend": "chat", + "endpoint": "http://100.72.9.84:3000/v1/chat/completions", + "model_requested": "frontier-infra/jebadiah-4b-v2", + "items_file": "bench/decide/items.json", + "item_set": "decide-110", + "items": 110, + "sets": [ + "route", + "triage", + "urgency", + "pr_safe", + "fact" + ], + "concurrency": 8, + "timeout_s": 120 + }, + "decide": { + "backend": "chat", + "endpoint": "http://100.72.9.84:3000/v1/chat/completions", + "model_reported": "frontier-infra/jebadiah-4b-v2", + "item_set": { + "id": "decide-110", + "version": 1, + "file": "items.json", + "count": 110, + "sets": { + "route": 30, + "triage": 20, + "urgency": 20, + "pr_safe": 20, + "fact": 20 + } + }, + "protocol": { + "backend": "chat", + "endpoint": "http://100.72.9.84:3000/v1/chat/completions", + "timeout_s": 120, + "input_usd_per_mtok": 0.0, + "output_usd_per_mtok": 0.0, + "concurrency": 8, + "bins": 5, + "thresholds": [ + "0.8", + "0.9" + ], + "confidence": "the probability the backend put on the answer it gave; its reported confidence when it gave no distribution", + "brier": "one term, on the labeled option's probability" + }, + "overall": { + "n": 110, + "answered": 110, + "errors": 0, + "accuracy": 0.973, + "brier": 0.027, + "ece": 0.08, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 2, + "accuracy": 0.5, + "confidence": 0.492 + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 16, + "accuracy": 0.875, + "confidence": 0.738 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 92, + "accuracy": 1.0, + "confidence": 0.929 + } + ], + "thresholds": { + "0.8": { + "kept": 92, + "wrong": 0, + "abstained": 18, + "no_confidence": 0 + }, + "0.9": { + "kept": 65, + "wrong": 0, + "abstained": 45, + "no_confidence": 0 + } + }, + "tokens": { + "in": 12275, + "out": 220 + }, + "cost_usd": 0.0, + "p50_ms": 492, + "p95_ms": 1722 + }, + "sets": { + "route": { + "n": 30, + "answered": 30, + "errors": 0, + "accuracy": 0.933, + "brier": 0.046, + "ece": 0.109, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 1, + "accuracy": 0.0, + "confidence": 0.421 + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 6, + "accuracy": 0.833, + "confidence": 0.729 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 23, + "accuracy": 1.0, + "confidence": 0.903 + } + ], + "thresholds": { + "0.8": { + "kept": 23, + "wrong": 0, + "abstained": 7, + "no_confidence": 0 + }, + "0.9": { + "kept": 11, + "wrong": 0, + "abstained": 19, + "no_confidence": 0 + } + }, + "tokens": { + "in": 3997, + "out": 60 + }, + "cost_usd": 0.0, + "p50_ms": 529, + "p95_ms": 1516 + }, + "triage": { + "n": 20, + "answered": 20, + "errors": 0, + "accuracy": 1.0, + "brier": 0.013, + "ece": 0.088, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 2, + "accuracy": 1.0, + "confidence": 0.743 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 18, + "accuracy": 1.0, + "confidence": 0.93 + } + ], + "thresholds": { + "0.8": { + "kept": 18, + "wrong": 0, + "abstained": 2, + "no_confidence": 0 + }, + "0.9": { + "kept": 14, + "wrong": 0, + "abstained": 6, + "no_confidence": 0 + } + }, + "tokens": { + "in": 2432, + "out": 40 + }, + "cost_usd": 0.0, + "p50_ms": 575, + "p95_ms": 1932 + }, + "urgency": { + "n": 20, + "answered": 20, + "errors": 0, + "accuracy": 0.95, + "brier": 0.033, + "ece": 0.063, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 3, + "accuracy": 0.667, + "confidence": 0.71 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 17, + "accuracy": 1.0, + "confidence": 0.933 + } + ], + "thresholds": { + "0.8": { + "kept": 17, + "wrong": 0, + "abstained": 3, + "no_confidence": 0 + }, + "0.9": { + "kept": 12, + "wrong": 0, + "abstained": 8, + "no_confidence": 0 + } + }, + "tokens": { + "in": 2132, + "out": 40 + }, + "cost_usd": 0.0, + "p50_ms": 474, + "p95_ms": 2427 + }, + "pr_safe": { + "n": 20, + "answered": 20, + "errors": 0, + "accuracy": 1.0, + "brier": 0.022, + "ece": 0.122, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 5, + "accuracy": 1.0, + "confidence": 0.763 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 15, + "accuracy": 1.0, + "confidence": 0.917 + } + ], + "thresholds": { + "0.8": { + "kept": 15, + "wrong": 0, + "abstained": 5, + "no_confidence": 0 + }, + "0.9": { + "kept": 10, + "wrong": 0, + "abstained": 10, + "no_confidence": 0 + } + }, + "tokens": { + "in": 2225, + "out": 40 + }, + "cost_usd": 0.0, + "p50_ms": 472, + "p95_ms": 1453 + }, + "fact": { + "n": 20, + "answered": 20, + "errors": 0, + "accuracy": 1.0, + "brier": 0.013, + "ece": 0.055, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 1, + "accuracy": 1.0, + "confidence": 0.562 + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 19, + "accuracy": 1.0, + "confidence": 0.965 + } + ], + "thresholds": { + "0.8": { + "kept": 19, + "wrong": 0, + "abstained": 1, + "no_confidence": 0 + }, + "0.9": { + "kept": 18, + "wrong": 0, + "abstained": 2, + "no_confidence": 0 + } + }, + "tokens": { + "in": 1489, + "out": 40 + }, + "cost_usd": 0.0, + "p50_ms": 423, + "p95_ms": 896 + } + }, + "rows": [ + { + "id": "route-01", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.929255, + "p_label": 0.929255, + "distribution": { + "code": 0.929255, + "chat": 0.067315, + "vision": 0.002033, + "long_document": 0.001397 + }, + "wall_ms": 696, + "server_latency_ms": null, + "tokens_in": 133, + "tokens_out": 2, + "error": null + }, + { + "id": "route-02", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.988709, + "p_label": 0.988709, + "distribution": { + "chat": 0.988709, + "vision": 0.005188, + "code": 0.003147, + "long_document": 0.002956 + }, + "wall_ms": 514, + "server_latency_ms": null, + "tokens_in": 132, + "tokens_out": 2, + "error": null + }, + { + "id": "route-03", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.851495, + "p_label": 0.851495, + "distribution": { + "vision": 0.851495, + "chat": 0.115237, + "long_document": 0.017672, + "code": 0.015596 + }, + "wall_ms": 384, + "server_latency_ms": null, + "tokens_in": 130, + "tokens_out": 2, + "error": null + }, + { + "id": "route-04", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.822619, + "p_label": 0.822619, + "distribution": { + "long_document": 0.822619, + "chat": 0.098248, + "vision": 0.076515, + "code": 0.002618 + }, + "wall_ms": 520, + "server_latency_ms": null, + "tokens_in": 139, + "tokens_out": 2, + "error": null + }, + { + "id": "route-05", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.668021, + "p_label": 0.668021, + "distribution": { + "code": 0.668021, + "chat": 0.315551, + "vision": 0.013864, + "long_document": 0.002565 + }, + "wall_ms": 1578, + "server_latency_ms": null, + "tokens_in": 138, + "tokens_out": 2, + "error": null + }, + { + "id": "route-06", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.985968, + "p_label": 0.985968, + "distribution": { + "chat": 0.985968, + "long_document": 0.007528, + "vision": 0.003556, + "code": 0.002948 + }, + "wall_ms": 514, + "server_latency_ms": null, + "tokens_in": 131, + "tokens_out": 2, + "error": null + }, + { + "id": "route-07", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.906211, + "p_label": 0.906211, + "distribution": { + "vision": 0.906211, + "chat": 0.074386, + "long_document": 0.014648, + "code": 0.004755 + }, + "wall_ms": 695, + "server_latency_ms": null, + "tokens_in": 134, + "tokens_out": 2, + "error": null + }, + { + "id": "route-08", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.822619, + "p_label": 0.822619, + "distribution": { + "long_document": 0.822619, + "chat": 0.098248, + "vision": 0.076515, + "code": 0.002618 + }, + "wall_ms": 512, + "server_latency_ms": null, + "tokens_in": 139, + "tokens_out": 2, + "error": null + }, + { + "id": "route-09", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.858495, + "p_label": 0.858495, + "distribution": { + "code": 0.858495, + "chat": 0.131654, + "vision": 0.006555, + "long_document": 0.003296 + }, + "wall_ms": 485, + "server_latency_ms": null, + "tokens_in": 133, + "tokens_out": 2, + "error": null + }, + { + "id": "route-10", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.989074, + "p_label": 0.989074, + "distribution": { + "chat": 0.989074, + "vision": 0.00519, + "code": 0.002957, + "long_document": 0.002778 + }, + "wall_ms": 472, + "server_latency_ms": null, + "tokens_in": 131, + "tokens_out": 2, + "error": null + }, + { + "id": "route-11", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.77479, + "p_label": 0.77479, + "distribution": { + "vision": 0.77479, + "chat": 0.195898, + "long_document": 0.023397, + "code": 0.005916 + }, + "wall_ms": 471, + "server_latency_ms": null, + "tokens_in": 134, + "tokens_out": 2, + "error": null + }, + { + "id": "route-12", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.855873, + "p_label": 0.855873, + "distribution": { + "long_document": 0.855873, + "vision": 0.061999, + "chat": 0.061999, + "code": 0.020128 + }, + "wall_ms": 470, + "server_latency_ms": null, + "tokens_in": 139, + "tokens_out": 2, + "error": null + }, + { + "id": "route-13", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.792855, + "p_label": 0.792855, + "distribution": { + "code": 0.792855, + "chat": 0.200465, + "vision": 0.004714, + "long_document": 0.001965 + }, + "wall_ms": 464, + "server_latency_ms": null, + "tokens_in": 135, + "tokens_out": 2, + "error": null + }, + { + "id": "route-14", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.989278, + "p_label": 0.989278, + "distribution": { + "chat": 0.989278, + "vision": 0.005191, + "code": 0.004043, + "long_document": 0.001487 + }, + "wall_ms": 441, + "server_latency_ms": null, + "tokens_in": 127, + "tokens_out": 2, + "error": null + }, + { + "id": "route-15", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.881346, + "p_label": 0.881346, + "distribution": { + "vision": 0.881346, + "chat": 0.105262, + "long_document": 0.009791, + "code": 0.003602 + }, + "wall_ms": 589, + "server_latency_ms": null, + "tokens_in": 129, + "tokens_out": 2, + "error": null + }, + { + "id": "route-16", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "vision", + "correct": false, + "p_answer": 0.421152, + "p_label": 0.421152, + "distribution": { + "vision": 0.421152, + "long_document": 0.421152, + "chat": 0.136728, + "code": 0.020968 + }, + "wall_ms": 417, + "server_latency_ms": null, + "tokens_in": 132, + "tokens_out": 2, + "error": null + }, + { + "id": "route-17", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "chat", + "correct": false, + "p_answer": 0.610571, + "p_label": 0.37033, + "distribution": { + "chat": 0.610571, + "code": 0.37033, + "vision": 0.016271, + "long_document": 0.002828 + }, + "wall_ms": 595, + "server_latency_ms": null, + "tokens_in": 133, + "tokens_out": 2, + "error": null + }, + { + "id": "route-18", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.984647, + "p_label": 0.984647, + "distribution": { + "chat": 0.984647, + "vision": 0.006635, + "code": 0.005167, + "long_document": 0.003551 + }, + "wall_ms": 604, + "server_latency_ms": null, + "tokens_in": 132, + "tokens_out": 2, + "error": null + }, + { + "id": "route-19", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.877865, + "p_label": 0.877865, + "distribution": { + "vision": 0.877865, + "code": 0.049526, + "long_document": 0.038571, + "chat": 0.034039 + }, + "wall_ms": 599, + "server_latency_ms": null, + "tokens_in": 131, + "tokens_out": 2, + "error": null + }, + { + "id": "route-20", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.812535, + "p_label": 0.812535, + "distribution": { + "long_document": 0.812535, + "chat": 0.159998, + "vision": 0.024536, + "code": 0.00293 + }, + "wall_ms": 599, + "server_latency_ms": null, + "tokens_in": 138, + "tokens_out": 2, + "error": null + }, + { + "id": "route-21", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.904748, + "p_label": 0.904748, + "distribution": { + "code": 0.904748, + "chat": 0.084155, + "vision": 0.006908, + "long_document": 0.00419 + }, + "wall_ms": 460, + "server_latency_ms": null, + "tokens_in": 127, + "tokens_out": 2, + "error": null + }, + { + "id": "route-22", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.989241, + "p_label": 0.989241, + "distribution": { + "chat": 0.989241, + "vision": 0.005191, + "code": 0.002958, + "long_document": 0.00261 + }, + "wall_ms": 1516, + "server_latency_ms": null, + "tokens_in": 130, + "tokens_out": 2, + "error": null + }, + { + "id": "route-23", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.827079, + "p_label": 0.827079, + "distribution": { + "vision": 0.827079, + "chat": 0.143725, + "long_document": 0.022041, + "code": 0.007156 + }, + "wall_ms": 894, + "server_latency_ms": null, + "tokens_in": 133, + "tokens_out": 2, + "error": null + }, + { + "id": "route-24", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.870018, + "p_label": 0.870018, + "distribution": { + "long_document": 0.870018, + "vision": 0.071415, + "chat": 0.055618, + "code": 0.002948 + }, + "wall_ms": 492, + "server_latency_ms": null, + "tokens_in": 141, + "tokens_out": 2, + "error": null + }, + { + "id": "route-25", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.813165, + "p_label": 0.813165, + "distribution": { + "code": 0.813165, + "chat": 0.181442, + "vision": 0.004009, + "long_document": 0.001385 + }, + "wall_ms": 547, + "server_latency_ms": null, + "tokens_in": 139, + "tokens_out": 2, + "error": null + }, + { + "id": "route-26", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.988508, + "p_label": 0.988508, + "distribution": { + "chat": 0.988508, + "vision": 0.005187, + "code": 0.003349, + "long_document": 0.002956 + }, + "wall_ms": 542, + "server_latency_ms": null, + "tokens_in": 129, + "tokens_out": 2, + "error": null + }, + { + "id": "route-27", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.82754, + "p_label": 0.82754, + "distribution": { + "vision": 0.82754, + "chat": 0.143805, + "long_document": 0.019462, + "code": 0.009193 + }, + "wall_ms": 1375, + "server_latency_ms": null, + "tokens_in": 132, + "tokens_out": 2, + "error": null + }, + { + "id": "route-28", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.755645, + "p_label": 0.755645, + "distribution": { + "long_document": 0.755645, + "vision": 0.131311, + "chat": 0.102265, + "code": 0.010779 + }, + "wall_ms": 537, + "server_latency_ms": null, + "tokens_in": 135, + "tokens_out": 2, + "error": null + }, + { + "id": "route-29", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.773829, + "p_label": 0.773829, + "distribution": { + "code": 0.773829, + "chat": 0.221706, + "vision": 0.002971, + "long_document": 0.001494 + }, + "wall_ms": 529, + "server_latency_ms": null, + "tokens_in": 129, + "tokens_out": 2, + "error": null + }, + { + "id": "route-30", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.986957, + "p_label": 0.986957, + "distribution": { + "chat": 0.986957, + "vision": 0.00665, + "code": 0.003789, + "long_document": 0.002604 + }, + "wall_ms": 632, + "server_latency_ms": null, + "tokens_in": 132, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-01", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.98892, + "p_label": 0.98892, + "distribution": { + "billing": 0.98892, + "sales": 0.00588, + "technical": 0.00458, + "account": 0.00062 + }, + "wall_ms": 577, + "server_latency_ms": null, + "tokens_in": 120, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-02", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.941269, + "p_label": 0.941269, + "distribution": { + "technical": 0.941269, + "billing": 0.053103, + "sales": 0.002814, + "account": 0.002814 + }, + "wall_ms": 544, + "server_latency_ms": null, + "tokens_in": 128, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-03", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.852747, + "p_label": 0.852747, + "distribution": { + "sales": 0.852747, + "billing": 0.130773, + "technical": 0.010734, + "account": 0.005746 + }, + "wall_ms": 575, + "server_latency_ms": null, + "tokens_in": 121, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-04", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.932898, + "p_label": 0.932898, + "distribution": { + "account": 0.932898, + "technical": 0.059638, + "sales": 0.005547, + "billing": 0.001917 + }, + "wall_ms": 575, + "server_latency_ms": null, + "tokens_in": 121, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-05", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.950728, + "p_label": 0.950728, + "distribution": { + "billing": 0.950728, + "account": 0.025336, + "sales": 0.011968, + "technical": 0.011968 + }, + "wall_ms": 522, + "server_latency_ms": null, + "tokens_in": 121, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-06", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.958517, + "p_label": 0.958517, + "distribution": { + "technical": 0.958517, + "billing": 0.022542, + "account": 0.010648, + "sales": 0.008293 + }, + "wall_ms": 401, + "server_latency_ms": null, + "tokens_in": 124, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-07", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.920108, + "p_label": 0.920108, + "distribution": { + "sales": 0.920108, + "billing": 0.058821, + "technical": 0.014872, + "account": 0.0062 + }, + "wall_ms": 2107, + "server_latency_ms": null, + "tokens_in": 121, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-08", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.930621, + "p_label": 0.930621, + "distribution": { + "account": 0.930621, + "sales": 0.046333, + "technical": 0.021886, + "billing": 0.00116 + }, + "wall_ms": 454, + "server_latency_ms": null, + "tokens_in": 120, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-09", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.885373, + "p_label": 0.885373, + "distribution": { + "billing": 0.885373, + "technical": 0.093318, + "sales": 0.012629, + "account": 0.00868 + }, + "wall_ms": 452, + "server_latency_ms": null, + "tokens_in": 120, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-10", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.911915, + "p_label": 0.911915, + "distribution": { + "technical": 0.911915, + "account": 0.045402, + "billing": 0.031204, + "sales": 0.011479 + }, + "wall_ms": 453, + "server_latency_ms": null, + "tokens_in": 123, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-11", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.827511, + "p_label": 0.827511, + "distribution": { + "sales": 0.827511, + "billing": 0.162946, + "technical": 0.004921, + "account": 0.004622 + }, + "wall_ms": 453, + "server_latency_ms": null, + "tokens_in": 117, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-12", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.943325, + "p_label": 0.943325, + "distribution": { + "account": 0.943325, + "technical": 0.036577, + "sales": 0.017278, + "billing": 0.00282 + }, + "wall_ms": 1932, + "server_latency_ms": null, + "tokens_in": 126, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-13", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.963311, + "p_label": 0.963311, + "distribution": { + "billing": 0.963311, + "sales": 0.032963, + "technical": 0.00198, + "account": 0.001747 + }, + "wall_ms": 1834, + "server_latency_ms": null, + "tokens_in": 118, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-14", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.973288, + "p_label": 0.973288, + "distribution": { + "technical": 0.973288, + "account": 0.017826, + "billing": 0.005787, + "sales": 0.003098 + }, + "wall_ms": 1722, + "server_latency_ms": null, + "tokens_in": 119, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-15", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.87142, + "p_label": 0.87142, + "distribution": { + "sales": 0.87142, + "technical": 0.104076, + "account": 0.015961, + "billing": 0.008543 + }, + "wall_ms": 1637, + "server_latency_ms": null, + "tokens_in": 120, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-16", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.738489, + "p_label": 0.738489, + "distribution": { + "account": 0.738489, + "technical": 0.239752, + "sales": 0.017368, + "billing": 0.004391 + }, + "wall_ms": 1639, + "server_latency_ms": null, + "tokens_in": 127, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-17", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.934925, + "p_label": 0.934925, + "distribution": { + "billing": 0.934925, + "sales": 0.059768, + "technical": 0.004067, + "account": 0.00124 + }, + "wall_ms": 1637, + "server_latency_ms": null, + "tokens_in": 122, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-18", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.970984, + "p_label": 0.970984, + "distribution": { + "technical": 0.970984, + "account": 0.022835, + "sales": 0.00309, + "billing": 0.00309 + }, + "wall_ms": 1637, + "server_latency_ms": null, + "tokens_in": 126, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-19", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.747254, + "p_label": 0.747254, + "distribution": { + "sales": 0.747254, + "billing": 0.242598, + "technical": 0.006465, + "account": 0.003684 + }, + "wall_ms": 317, + "server_latency_ms": null, + "tokens_in": 119, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-20", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.98727, + "p_label": 0.98727, + "distribution": { + "account": 0.98727, + "sales": 0.008542, + "billing": 0.00216, + "technical": 0.002029 + }, + "wall_ms": 534, + "server_latency_ms": null, + "tokens_in": 119, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-01", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.731059, + "p_label": 0.731059, + "distribution": { + "true": 0.731059, + "false": 0.268941 + }, + "wall_ms": 2428, + "server_latency_ms": null, + "tokens_in": 105, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-02", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.991423, + "p_label": 0.991423, + "distribution": { + "true": 0.991423, + "false": 0.008577 + }, + "wall_ms": 533, + "server_latency_ms": null, + "tokens_in": 113, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-03", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.977023, + "p_label": 0.977023, + "distribution": { + "false": 0.977023, + "true": 0.022977 + }, + "wall_ms": 2427, + "server_latency_ms": null, + "tokens_in": 106, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-04", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": false, + "correct": false, + "p_answer": 0.622459, + "p_label": 0.377541, + "distribution": { + "false": 0.622459, + "true": 0.377541 + }, + "wall_ms": 532, + "server_latency_ms": null, + "tokens_in": 106, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-05", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.835484, + "p_label": 0.835484, + "distribution": { + "false": 0.835484, + "true": 0.164516 + }, + "wall_ms": 532, + "server_latency_ms": null, + "tokens_in": 106, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-06", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.984094, + "p_label": 0.984094, + "distribution": { + "false": 0.984094, + "true": 0.015906 + }, + "wall_ms": 517, + "server_latency_ms": null, + "tokens_in": 109, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-07", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.985936, + "p_label": 0.985936, + "distribution": { + "false": 0.985936, + "true": 0.014064 + }, + "wall_ms": 355, + "server_latency_ms": null, + "tokens_in": 106, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-08", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.939913, + "p_label": 0.939913, + "distribution": { + "false": 0.939913, + "true": 0.060087 + }, + "wall_ms": 403, + "server_latency_ms": null, + "tokens_in": 105, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-09", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.952574, + "p_label": 0.952574, + "distribution": { + "true": 0.952574, + "false": 0.047426 + }, + "wall_ms": 405, + "server_latency_ms": null, + "tokens_in": 105, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-10", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.835484, + "p_label": 0.835484, + "distribution": { + "false": 0.835484, + "true": 0.164516 + }, + "wall_ms": 473, + "server_latency_ms": null, + "tokens_in": 108, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-11", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.987568, + "p_label": 0.987568, + "distribution": { + "false": 0.987568, + "true": 0.012432 + }, + "wall_ms": 474, + "server_latency_ms": null, + "tokens_in": 102, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-12", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.904651, + "p_label": 0.904651, + "distribution": { + "true": 0.904651, + "false": 0.095349 + }, + "wall_ms": 472, + "server_latency_ms": null, + "tokens_in": 111, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-13", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.835484, + "p_label": 0.835484, + "distribution": { + "false": 0.835484, + "true": 0.164516 + }, + "wall_ms": 394, + "server_latency_ms": null, + "tokens_in": 103, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-14", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.817574, + "p_label": 0.817574, + "distribution": { + "true": 0.817574, + "false": 0.182426 + }, + "wall_ms": 305, + "server_latency_ms": null, + "tokens_in": 104, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-15", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.982014, + "p_label": 0.982014, + "distribution": { + "false": 0.982014, + "true": 0.017986 + }, + "wall_ms": 1622, + "server_latency_ms": null, + "tokens_in": 105, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-16", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.987568, + "p_label": 0.987568, + "distribution": { + "true": 0.987568, + "false": 0.012432 + }, + "wall_ms": 404, + "server_latency_ms": null, + "tokens_in": 112, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-17", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.7773, + "p_label": 0.7773, + "distribution": { + "false": 0.7773, + "true": 0.2227 + }, + "wall_ms": 558, + "server_latency_ms": null, + "tokens_in": 107, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-18", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.893309, + "p_label": 0.893309, + "distribution": { + "true": 0.893309, + "false": 0.106691 + }, + "wall_ms": 404, + "server_latency_ms": null, + "tokens_in": 111, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-19", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.977023, + "p_label": 0.977023, + "distribution": { + "false": 0.977023, + "true": 0.022977 + }, + "wall_ms": 518, + "server_latency_ms": null, + "tokens_in": 104, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-20", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.974043, + "p_label": 0.974043, + "distribution": { + "false": 0.974043, + "true": 0.025957 + }, + "wall_ms": 1453, + "server_latency_ms": null, + "tokens_in": 104, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-01", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.924142, + "p_label": 0.924142, + "distribution": { + "true": 0.924142, + "false": 0.075858 + }, + "wall_ms": 425, + "server_latency_ms": null, + "tokens_in": 114, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-02", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.974043, + "p_label": 0.974043, + "distribution": { + "true": 0.974043, + "false": 0.025957 + }, + "wall_ms": 409, + "server_latency_ms": null, + "tokens_in": 107, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-03", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.7773, + "p_label": 0.7773, + "distribution": { + "false": 0.7773, + "true": 0.2227 + }, + "wall_ms": 388, + "server_latency_ms": null, + "tokens_in": 107, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-04", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.979668, + "p_label": 0.979668, + "distribution": { + "false": 0.979668, + "true": 0.020332 + }, + "wall_ms": 367, + "server_latency_ms": null, + "tokens_in": 107, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-05", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.851953, + "p_label": 0.851953, + "distribution": { + "true": 0.851953, + "false": 0.148047 + }, + "wall_ms": 320, + "server_latency_ms": null, + "tokens_in": 119, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-06", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.798187, + "p_label": 0.798187, + "distribution": { + "false": 0.798187, + "true": 0.201813 + }, + "wall_ms": 303, + "server_latency_ms": null, + "tokens_in": 113, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-07", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.992423, + "p_label": 0.992423, + "distribution": { + "true": 0.992423, + "false": 0.007577 + }, + "wall_ms": 988, + "server_latency_ms": null, + "tokens_in": 109, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-08", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.962673, + "p_label": 0.962673, + "distribution": { + "false": 0.962673, + "true": 0.037327 + }, + "wall_ms": 680, + "server_latency_ms": null, + "tokens_in": 110, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-09", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.952574, + "p_label": 0.952574, + "distribution": { + "true": 0.952574, + "false": 0.047426 + }, + "wall_ms": 474, + "server_latency_ms": null, + "tokens_in": 107, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-10", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.754915, + "p_label": 0.754915, + "distribution": { + "false": 0.754915, + "true": 0.245085 + }, + "wall_ms": 472, + "server_latency_ms": null, + "tokens_in": 112, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-11", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.952574, + "p_label": 0.952574, + "distribution": { + "true": 0.952574, + "false": 0.047426 + }, + "wall_ms": 474, + "server_latency_ms": null, + "tokens_in": 113, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-12", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.817574, + "p_label": 0.817574, + "distribution": { + "false": 0.817574, + "true": 0.182426 + }, + "wall_ms": 474, + "server_latency_ms": null, + "tokens_in": 116, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-13", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.932453, + "p_label": 0.932453, + "distribution": { + "true": 0.932453, + "false": 0.067547 + }, + "wall_ms": 444, + "server_latency_ms": null, + "tokens_in": 112, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-14", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.932453, + "p_label": 0.932453, + "distribution": { + "false": 0.932453, + "true": 0.067547 + }, + "wall_ms": 437, + "server_latency_ms": null, + "tokens_in": 112, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-15", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.754915, + "p_label": 0.754915, + "distribution": { + "true": 0.754915, + "false": 0.245085 + }, + "wall_ms": 488, + "server_latency_ms": null, + "tokens_in": 110, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-16", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.851953, + "p_label": 0.851953, + "distribution": { + "false": 0.851953, + "true": 0.148047 + }, + "wall_ms": 488, + "server_latency_ms": null, + "tokens_in": 119, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-17", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.851953, + "p_label": 0.851953, + "distribution": { + "true": 0.851953, + "false": 0.148047 + }, + "wall_ms": 354, + "server_latency_ms": null, + "tokens_in": 109, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-18", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.731059, + "p_label": 0.731059, + "distribution": { + "false": 0.731059, + "true": 0.268941 + }, + "wall_ms": 1506, + "server_latency_ms": null, + "tokens_in": 108, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-19", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.939913, + "p_label": 0.939913, + "distribution": { + "true": 0.939913, + "false": 0.060087 + }, + "wall_ms": 1453, + "server_latency_ms": null, + "tokens_in": 107, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-20", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.835484, + "p_label": 0.835484, + "distribution": { + "false": 0.835484, + "true": 0.164516 + }, + "wall_ms": 1422, + "server_latency_ms": null, + "tokens_in": 114, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-01", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.991423, + "p_label": 0.991423, + "distribution": { + "true": 0.991423, + "false": 0.008577 + }, + "wall_ms": 818, + "server_latency_ms": null, + "tokens_in": 76, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-02", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.924142, + "p_label": 0.924142, + "distribution": { + "false": 0.924142, + "true": 0.075858 + }, + "wall_ms": 887, + "server_latency_ms": null, + "tokens_in": 78, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-03", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.999643, + "p_label": 0.999643, + "distribution": { + "true": 0.999643, + "false": 0.000357 + }, + "wall_ms": 843, + "server_latency_ms": null, + "tokens_in": 68, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-04", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.932453, + "p_label": 0.932453, + "distribution": { + "true": 0.932453, + "false": 0.067547 + }, + "wall_ms": 896, + "server_latency_ms": null, + "tokens_in": 70, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-05", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.562177, + "p_label": 0.562177, + "distribution": { + "true": 0.562177, + "false": 0.437823 + }, + "wall_ms": 791, + "server_latency_ms": null, + "tokens_in": 83, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-06", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.977023, + "p_label": 0.977023, + "distribution": { + "false": 0.977023, + "true": 0.022977 + }, + "wall_ms": 1347, + "server_latency_ms": null, + "tokens_in": 77, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-07", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.914901, + "p_label": 0.914901, + "distribution": { + "false": 0.914901, + "true": 0.085099 + }, + "wall_ms": 434, + "server_latency_ms": null, + "tokens_in": 74, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-08", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.99929, + "p_label": 0.99929, + "distribution": { + "true": 0.99929, + "false": 0.00071 + }, + "wall_ms": 433, + "server_latency_ms": null, + "tokens_in": 77, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-09", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.999089, + "p_label": 0.999089, + "distribution": { + "true": 0.999089, + "false": 0.000911 + }, + "wall_ms": 421, + "server_latency_ms": null, + "tokens_in": 70, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-10", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.939913, + "p_label": 0.939913, + "distribution": { + "false": 0.939913, + "true": 0.060087 + }, + "wall_ms": 360, + "server_latency_ms": null, + "tokens_in": 74, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-11", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.99883, + "p_label": 0.99883, + "distribution": { + "true": 0.99883, + "false": 0.00117 + }, + "wall_ms": 398, + "server_latency_ms": null, + "tokens_in": 82, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-12", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.994089, + "p_label": 0.994089, + "distribution": { + "true": 0.994089, + "false": 0.005911 + }, + "wall_ms": 398, + "server_latency_ms": null, + "tokens_in": 74, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-13", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.992423, + "p_label": 0.992423, + "distribution": { + "false": 0.992423, + "true": 0.007577 + }, + "wall_ms": 394, + "server_latency_ms": null, + "tokens_in": 71, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-14", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.99903, + "p_label": 0.99903, + "distribution": { + "true": 0.99903, + "false": 0.00097 + }, + "wall_ms": 431, + "server_latency_ms": null, + "tokens_in": 73, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-15", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.932453, + "p_label": 0.932453, + "distribution": { + "false": 0.932453, + "true": 0.067547 + }, + "wall_ms": 424, + "server_latency_ms": null, + "tokens_in": 71, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-16", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.817574, + "p_label": 0.817574, + "distribution": { + "true": 0.817574, + "false": 0.182426 + }, + "wall_ms": 423, + "server_latency_ms": null, + "tokens_in": 79, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-17", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.989013, + "p_label": 0.989013, + "distribution": { + "false": 0.989013, + "true": 0.010987 + }, + "wall_ms": 371, + "server_latency_ms": null, + "tokens_in": 72, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-18", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.952574, + "p_label": 0.952574, + "distribution": { + "true": 0.952574, + "false": 0.047426 + }, + "wall_ms": 372, + "server_latency_ms": null, + "tokens_in": 80, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-19", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.982014, + "p_label": 0.982014, + "distribution": { + "false": 0.982014, + "true": 0.017986 + }, + "wall_ms": 414, + "server_latency_ms": null, + "tokens_in": 69, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-20", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.999089, + "p_label": 0.999089, + "distribution": { + "true": 0.999089, + "false": 0.000911 + }, + "wall_ms": 414, + "server_latency_ms": null, + "tokens_in": 71, + "tokens_out": 2, + "error": null + } + ] + }, + "notes": [ + "Measured by scripts/ainode-bench.py decide in 10s over 110 labeled items from items.json (decide-110); nothing was loaded, unloaded or restarted.", + "Accuracy is the weakest number here: the failure mode automation cares about is a wrong answer at high confidence, which is what the Brier score, the calibration error and the wrong-at-threshold counts measure.", + "Confidence is the probability the backend put on the answer it gave. Brier is a single term on the labeled option's probability, and the calibration error uses 5 bins over the reliability table in the same block.", + "Cost is $0 for the chat backend: it runs on our own hardware and nobody bills per token for it. The electricity is real and is not a number this record claims to have measured.", + "The chat backend is the fallback instrument: lettered options with thinking off, and a distribution softmaxed from the top logprobs of the one letter token. Those probabilities are over letters rather than over meanings, so they are a weaker statement than a decision endpoint's.", + "Spark-4-GX10 does not report serving frontier-infra/jebadiah-4b-v2; placement and telemetry may describe the wrong node", + "frontier-infra/jebadiah-4b-v2 is not in the AINode catalog; fill model metadata (params_b, license, context) by hand" + ], + "source": "scripts/ainode-bench.py decide" +} diff --git a/bench/results/20261002-005114-nvidia-nemotron-3-nano-4b-bf16-spark-4-stacked-beside-jeb-4b-and-9b-vllm-0_27_1-bf16-decide.json b/bench/results/20261002-005114-nvidia-nemotron-3-nano-4b-bf16-spark-4-stacked-beside-jeb-4b-and-9b-vllm-0_27_1-bf16-decide.json new file mode 100644 index 00000000..bed22a44 --- /dev/null +++ b/bench/results/20261002-005114-nvidia-nemotron-3-nano-4b-bf16-spark-4-stacked-beside-jeb-4b-and-9b-vllm-0_27_1-bf16-decide.json @@ -0,0 +1,2675 @@ +{ + "schema": 1, + "stamp": "20261002-005114", + "label": "spark-4 stacked beside jeb 4b and 9b, vllm 0.27.1 bf16", + "model": { + "id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16", + "name": "NVIDIA Nemotron 3 Nano 4B BF16", + "params_b": 4.0, + "vision": false, + "arch": "dense", + "active_b": 4.0, + "quant": "BF16" + }, + "placement": { + "engine": "vllm", + "node": "Spark-4-GX10", + "gpu": "NVIDIA GB10", + "ainode": "0.5.33", + "tp": 1, + "gpus": 1, + "stacked_with": [ + "frontier-infra/jebadiah-9b-v2", + "frontier-infra/jebadiah-4b-v2" + ], + "port": 8000 + }, + "settings": { + "backend": "chat", + "endpoint": "http://100.72.9.84:3000/v1/chat/completions", + "model_requested": "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16", + "items_file": "bench/decide/items.json", + "item_set": "decide-110", + "items": 110, + "sets": [ + "route", + "triage", + "urgency", + "pr_safe", + "fact" + ], + "concurrency": 8, + "timeout_s": 120 + }, + "decide": { + "backend": "chat", + "endpoint": "http://100.72.9.84:3000/v1/chat/completions", + "model_reported": "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16", + "item_set": { + "id": "decide-110", + "version": 1, + "file": "items.json", + "count": 110, + "sets": { + "route": 30, + "triage": 20, + "urgency": 20, + "pr_safe": 20, + "fact": 20 + } + }, + "protocol": { + "backend": "chat", + "endpoint": "http://100.72.9.84:3000/v1/chat/completions", + "timeout_s": 120, + "input_usd_per_mtok": 0.0, + "output_usd_per_mtok": 0.0, + "concurrency": 8, + "bins": 5, + "thresholds": [ + "0.8", + "0.9" + ], + "confidence": "the probability the backend put on the answer it gave; its reported confidence when it gave no distribution", + "brier": "one term, on the labeled option's probability" + }, + "overall": { + "n": 110, + "answered": 110, + "errors": 0, + "accuracy": 0.945, + "brier": 0.054, + "ece": 0.045, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 8, + "accuracy": 0.75, + "confidence": 0.57 + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 14, + "accuracy": 0.929, + "confidence": 0.726 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 88, + "accuracy": 0.966, + "confidence": 0.973 + } + ], + "thresholds": { + "0.8": { + "kept": 88, + "wrong": 3, + "abstained": 22, + "no_confidence": 0 + }, + "0.9": { + "kept": 80, + "wrong": 1, + "abstained": 30, + "no_confidence": 0 + } + }, + "tokens": { + "in": 12041, + "out": 220 + }, + "cost_usd": 0.0, + "p50_ms": 407, + "p95_ms": 1696 + }, + "sets": { + "route": { + "n": 30, + "answered": 30, + "errors": 0, + "accuracy": 0.967, + "brier": 0.04, + "ece": 0.05, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 1, + "accuracy": 1.0, + "confidence": 0.561 + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 3, + "accuracy": 1.0, + "confidence": 0.737 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 26, + "accuracy": 0.962, + "confidence": 0.972 + } + ], + "thresholds": { + "0.8": { + "kept": 26, + "wrong": 1, + "abstained": 4, + "no_confidence": 0 + }, + "0.9": { + "kept": 23, + "wrong": 0, + "abstained": 7, + "no_confidence": 0 + } + }, + "tokens": { + "in": 3884, + "out": 60 + }, + "cost_usd": 0.0, + "p50_ms": 471, + "p95_ms": 1697 + }, + "triage": { + "n": 20, + "answered": 20, + "errors": 0, + "accuracy": 1.0, + "brier": 0.008, + "ece": 0.041, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 2, + "accuracy": 1.0, + "confidence": 0.747 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 18, + "accuracy": 1.0, + "confidence": 0.982 + } + ], + "thresholds": { + "0.8": { + "kept": 18, + "wrong": 0, + "abstained": 2, + "no_confidence": 0 + }, + "0.9": { + "kept": 17, + "wrong": 0, + "abstained": 3, + "no_confidence": 0 + } + }, + "tokens": { + "in": 2455, + "out": 40 + }, + "cost_usd": 0.0, + "p50_ms": 409, + "p95_ms": 448 + }, + "urgency": { + "n": 20, + "answered": 20, + "errors": 0, + "accuracy": 0.85, + "brier": 0.116, + "ece": 0.094, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 2, + "accuracy": 0.5, + "confidence": 0.562 + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 1, + "accuracy": 1.0, + "confidence": 0.777 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 17, + "accuracy": 0.882, + "confidence": 0.973 + } + ], + "thresholds": { + "0.8": { + "kept": 17, + "wrong": 2, + "abstained": 3, + "no_confidence": 0 + }, + "0.9": { + "kept": 16, + "wrong": 1, + "abstained": 4, + "no_confidence": 0 + } + }, + "tokens": { + "in": 2075, + "out": 40 + }, + "cost_usd": 0.0, + "p50_ms": 398, + "p95_ms": 403 + }, + "pr_safe": { + "n": 20, + "answered": 20, + "errors": 0, + "accuracy": 0.95, + "brier": 0.067, + "ece": 0.107, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 2, + "accuracy": 1.0, + "confidence": 0.562 + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 6, + "accuracy": 0.833, + "confidence": 0.721 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 12, + "accuracy": 1.0, + "confidence": 0.95 + } + ], + "thresholds": { + "0.8": { + "kept": 12, + "wrong": 0, + "abstained": 8, + "no_confidence": 0 + }, + "0.9": { + "kept": 9, + "wrong": 0, + "abstained": 11, + "no_confidence": 0 + } + }, + "tokens": { + "in": 2215, + "out": 40 + }, + "cost_usd": 0.0, + "p50_ms": 406, + "p95_ms": 411 + }, + "fact": { + "n": 20, + "answered": 20, + "errors": 0, + "accuracy": 0.95, + "brier": 0.043, + "ece": 0.057, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 3, + "accuracy": 0.667, + "confidence": 0.583 + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 2, + "accuracy": 1.0, + "confidence": 0.679 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 15, + "accuracy": 1.0, + "confidence": 0.983 + } + ], + "thresholds": { + "0.8": { + "kept": 15, + "wrong": 0, + "abstained": 5, + "no_confidence": 0 + }, + "0.9": { + "kept": 15, + "wrong": 0, + "abstained": 5, + "no_confidence": 0 + } + }, + "tokens": { + "in": 1412, + "out": 40 + }, + "cost_usd": 0.0, + "p50_ms": 361, + "p95_ms": 375 + } + }, + "rows": [ + { + "id": "route-01", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.982123, + "p_label": 0.982123, + "distribution": { + "code": 0.982123, + "chat": 0.015875, + "vision": 0.001896, + "long_document": 0.000107 + }, + "wall_ms": 1696, + "server_latency_ms": null, + "tokens_in": 130, + "tokens_out": 2, + "error": null + }, + { + "id": "route-02", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.998696, + "p_label": 0.998696, + "distribution": { + "chat": 0.998696, + "vision": 0.000911, + "code": 0.000335, + "long_document": 5.8e-05 + }, + "wall_ms": 1697, + "server_latency_ms": null, + "tokens_in": 128, + "tokens_out": 2, + "error": null + }, + { + "id": "route-03", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.997871, + "p_label": 0.997871, + "distribution": { + "vision": 0.997871, + "chat": 0.001324, + "code": 0.000625, + "long_document": 0.000179 + }, + "wall_ms": 1696, + "server_latency_ms": null, + "tokens_in": 126, + "tokens_out": 2, + "error": null + }, + { + "id": "route-04", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.985714, + "p_label": 0.985714, + "distribution": { + "long_document": 0.985714, + "chat": 0.007526, + "vision": 0.005861, + "code": 0.000899 + }, + "wall_ms": 1696, + "server_latency_ms": null, + "tokens_in": 136, + "tokens_out": 2, + "error": null + }, + { + "id": "route-05", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.842329, + "p_label": 0.842329, + "distribution": { + "code": 0.842329, + "chat": 0.146375, + "vision": 0.008258, + "long_document": 0.003038 + }, + "wall_ms": 1697, + "server_latency_ms": null, + "tokens_in": 134, + "tokens_out": 2, + "error": null + }, + { + "id": "route-06", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.999403, + "p_label": 0.999403, + "distribution": { + "chat": 0.999403, + "vision": 0.00038, + "code": 0.000203, + "long_document": 1.4e-05 + }, + "wall_ms": 1697, + "server_latency_ms": null, + "tokens_in": 127, + "tokens_out": 2, + "error": null + }, + { + "id": "route-07", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.992221, + "p_label": 0.992221, + "distribution": { + "vision": 0.992221, + "chat": 0.003579, + "code": 0.003579, + "long_document": 0.000622 + }, + "wall_ms": 1696, + "server_latency_ms": null, + "tokens_in": 130, + "tokens_out": 2, + "error": null + }, + { + "id": "route-08", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.996934, + "p_label": 0.996934, + "distribution": { + "long_document": 0.996934, + "chat": 0.002471, + "code": 0.000334, + "vision": 0.00026 + }, + "wall_ms": 1696, + "server_latency_ms": null, + "tokens_in": 135, + "tokens_out": 2, + "error": null + }, + { + "id": "route-09", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.936478, + "p_label": 0.936478, + "distribution": { + "code": 0.936478, + "chat": 0.059867, + "vision": 0.003377, + "long_document": 0.000277 + }, + "wall_ms": 441, + "server_latency_ms": null, + "tokens_in": 129, + "tokens_out": 2, + "error": null + }, + { + "id": "route-10", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.999201, + "p_label": 0.999201, + "distribution": { + "chat": 0.999201, + "vision": 0.000488, + "code": 0.000296, + "long_document": 1.6e-05 + }, + "wall_ms": 481, + "server_latency_ms": null, + "tokens_in": 127, + "tokens_out": 2, + "error": null + }, + { + "id": "route-11", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.899309, + "p_label": 0.899309, + "distribution": { + "vision": 0.899309, + "chat": 0.094786, + "code": 0.003675, + "long_document": 0.002229 + }, + "wall_ms": 481, + "server_latency_ms": null, + "tokens_in": 130, + "tokens_out": 2, + "error": null + }, + { + "id": "route-12", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.994117, + "p_label": 0.994117, + "distribution": { + "long_document": 0.994117, + "code": 0.003164, + "chat": 0.001919, + "vision": 0.0008 + }, + "wall_ms": 478, + "server_latency_ms": null, + "tokens_in": 136, + "tokens_out": 2, + "error": null + }, + { + "id": "route-13", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.676852, + "p_label": 0.676852, + "distribution": { + "code": 0.676852, + "chat": 0.319722, + "vision": 0.003134, + "long_document": 0.000292 + }, + "wall_ms": 479, + "server_latency_ms": null, + "tokens_in": 131, + "tokens_out": 2, + "error": null + }, + { + "id": "route-14", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.99455, + "p_label": 0.99455, + "distribution": { + "chat": 0.99455, + "code": 0.004064, + "vision": 0.00132, + "long_document": 6.6e-05 + }, + "wall_ms": 479, + "server_latency_ms": null, + "tokens_in": 123, + "tokens_out": 2, + "error": null + }, + { + "id": "route-15", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.979066, + "p_label": 0.979066, + "distribution": { + "vision": 0.979066, + "chat": 0.013966, + "code": 0.005822, + "long_document": 0.001146 + }, + "wall_ms": 481, + "server_latency_ms": null, + "tokens_in": 125, + "tokens_out": 2, + "error": null + }, + { + "id": "route-16", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.749766, + "p_label": 0.749766, + "distribution": { + "long_document": 0.749766, + "chat": 0.10147, + "vision": 0.079025, + "code": 0.069739 + }, + "wall_ms": 478, + "server_latency_ms": null, + "tokens_in": 128, + "tokens_out": 2, + "error": null + }, + { + "id": "route-17", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "chat", + "correct": false, + "p_answer": 0.865221, + "p_label": 0.132686, + "distribution": { + "chat": 0.865221, + "code": 0.132686, + "vision": 0.00167, + "long_document": 0.000422 + }, + "wall_ms": 471, + "server_latency_ms": null, + "tokens_in": 129, + "tokens_out": 2, + "error": null + }, + { + "id": "route-18", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.995935, + "p_label": 0.995935, + "distribution": { + "chat": 0.995935, + "vision": 0.002797, + "code": 0.001166, + "long_document": 0.000102 + }, + "wall_ms": 468, + "server_latency_ms": null, + "tokens_in": 128, + "tokens_out": 2, + "error": null + }, + { + "id": "route-19", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.983036, + "p_label": 0.983036, + "distribution": { + "vision": 0.983036, + "code": 0.014022, + "chat": 0.00215, + "long_document": 0.000791 + }, + "wall_ms": 467, + "server_latency_ms": null, + "tokens_in": 127, + "tokens_out": 2, + "error": null + }, + { + "id": "route-20", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.974332, + "p_label": 0.974332, + "distribution": { + "long_document": 0.974332, + "chat": 0.020222, + "code": 0.003982, + "vision": 0.001465 + }, + "wall_ms": 465, + "server_latency_ms": null, + "tokens_in": 134, + "tokens_out": 2, + "error": null + }, + { + "id": "route-21", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.97559, + "p_label": 0.97559, + "distribution": { + "code": 0.97559, + "chat": 0.020248, + "vision": 0.003987, + "long_document": 0.000175 + }, + "wall_ms": 466, + "server_latency_ms": null, + "tokens_in": 124, + "tokens_out": 2, + "error": null + }, + { + "id": "route-22", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.998151, + "p_label": 0.998151, + "distribution": { + "chat": 0.998151, + "vision": 0.001324, + "code": 0.000487, + "long_document": 3.8e-05 + }, + "wall_ms": 464, + "server_latency_ms": null, + "tokens_in": 126, + "tokens_out": 2, + "error": null + }, + { + "id": "route-23", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.985152, + "p_label": 0.985152, + "distribution": { + "vision": 0.985152, + "chat": 0.012401, + "code": 0.001902, + "long_document": 0.000545 + }, + "wall_ms": 464, + "server_latency_ms": null, + "tokens_in": 129, + "tokens_out": 2, + "error": null + }, + { + "id": "route-24", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.997422, + "p_label": 0.997422, + "distribution": { + "long_document": 0.997422, + "vision": 0.001168, + "chat": 0.001031, + "code": 0.000379 + }, + "wall_ms": 463, + "server_latency_ms": null, + "tokens_in": 137, + "tokens_out": 2, + "error": null + }, + { + "id": "route-25", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.560789, + "p_label": 0.560789, + "distribution": { + "code": 0.560789, + "chat": 0.436743, + "vision": 0.002292, + "long_document": 0.000177 + }, + "wall_ms": 444, + "server_latency_ms": null, + "tokens_in": 134, + "tokens_out": 2, + "error": null + }, + { + "id": "route-26", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.996888, + "p_label": 0.996888, + "distribution": { + "chat": 0.996888, + "vision": 0.001924, + "code": 0.00103, + "long_document": 0.000158 + }, + "wall_ms": 451, + "server_latency_ms": null, + "tokens_in": 126, + "tokens_out": 2, + "error": null + }, + { + "id": "route-27", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.991811, + "p_label": 0.991811, + "distribution": { + "vision": 0.991811, + "chat": 0.005898, + "code": 0.001915, + "long_document": 0.000377 + }, + "wall_ms": 450, + "server_latency_ms": null, + "tokens_in": 128, + "tokens_out": 2, + "error": null + }, + { + "id": "route-28", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.784343, + "p_label": 0.784343, + "distribution": { + "long_document": 0.784343, + "vision": 0.106149, + "chat": 0.082669, + "code": 0.026839 + }, + "wall_ms": 450, + "server_latency_ms": null, + "tokens_in": 132, + "tokens_out": 2, + "error": null + }, + { + "id": "route-29", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.91953, + "p_label": 0.91953, + "distribution": { + "code": 0.91953, + "chat": 0.07548, + "vision": 0.004825, + "long_document": 0.000165 + }, + "wall_ms": 450, + "server_latency_ms": null, + "tokens_in": 125, + "tokens_out": 2, + "error": null + }, + { + "id": "route-30", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.995374, + "p_label": 0.995374, + "distribution": { + "chat": 0.995374, + "vision": 0.002796, + "code": 0.001496, + "long_document": 0.000334 + }, + "wall_ms": 450, + "server_latency_ms": null, + "tokens_in": 130, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-01", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.999096, + "p_label": 0.999096, + "distribution": { + "billing": 0.999096, + "technical": 0.000804, + "sales": 9.6e-05, + "account": 3e-06 + }, + "wall_ms": 448, + "server_latency_ms": null, + "tokens_in": 121, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-02", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.999565, + "p_label": 0.999565, + "distribution": { + "technical": 0.999565, + "billing": 0.000261, + "sales": 0.000158, + "account": 1.6e-05 + }, + "wall_ms": 450, + "server_latency_ms": null, + "tokens_in": 129, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-03", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.911324, + "p_label": 0.911324, + "distribution": { + "sales": 0.911324, + "billing": 0.084766, + "technical": 0.003724, + "account": 0.000185 + }, + "wall_ms": 398, + "server_latency_ms": null, + "tokens_in": 122, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-04", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.996536, + "p_label": 0.996536, + "distribution": { + "account": 0.996536, + "technical": 0.002799, + "sales": 0.000486, + "billing": 0.000179 + }, + "wall_ms": 408, + "server_latency_ms": null, + "tokens_in": 122, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-05", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.992727, + "p_label": 0.992727, + "distribution": { + "billing": 0.992727, + "technical": 0.006689, + "sales": 0.000549, + "account": 3.5e-05 + }, + "wall_ms": 407, + "server_latency_ms": null, + "tokens_in": 122, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-06", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.999186, + "p_label": 0.999186, + "distribution": { + "technical": 0.999186, + "sales": 0.000488, + "billing": 0.00023, + "account": 9.6e-05 + }, + "wall_ms": 409, + "server_latency_ms": null, + "tokens_in": 125, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-07", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.988837, + "p_label": 0.988837, + "distribution": { + "sales": 0.988837, + "billing": 0.008555, + "technical": 0.002451, + "account": 0.000157 + }, + "wall_ms": 409, + "server_latency_ms": null, + "tokens_in": 123, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-08", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.992888, + "p_label": 0.992888, + "distribution": { + "account": 0.992888, + "sales": 0.00316, + "technical": 0.002789, + "billing": 0.001163 + }, + "wall_ms": 408, + "server_latency_ms": null, + "tokens_in": 121, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-09", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.968869, + "p_label": 0.968869, + "distribution": { + "billing": 0.968869, + "technical": 0.029257, + "sales": 0.001457, + "account": 0.000417 + }, + "wall_ms": 409, + "server_latency_ms": null, + "tokens_in": 121, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-10", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.995575, + "p_label": 0.995575, + "distribution": { + "technical": 0.995575, + "sales": 0.002178, + "billing": 0.001696, + "account": 0.000551 + }, + "wall_ms": 407, + "server_latency_ms": null, + "tokens_in": 124, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-11", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.877042, + "p_label": 0.877042, + "distribution": { + "sales": 0.877042, + "billing": 0.118695, + "technical": 0.004062, + "account": 0.000202 + }, + "wall_ms": 428, + "server_latency_ms": null, + "tokens_in": 120, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-12", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.999311, + "p_label": 0.999311, + "distribution": { + "account": 0.999311, + "technical": 0.000335, + "sales": 0.000296, + "billing": 5.8e-05 + }, + "wall_ms": 436, + "server_latency_ms": null, + "tokens_in": 127, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-13", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.995037, + "p_label": 0.995037, + "distribution": { + "billing": 0.995037, + "technical": 0.003589, + "sales": 0.00132, + "account": 5.4e-05 + }, + "wall_ms": 433, + "server_latency_ms": null, + "tokens_in": 119, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-14", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.99961, + "p_label": 0.99961, + "distribution": { + "technical": 0.99961, + "sales": 0.000203, + "account": 9.6e-05, + "billing": 9e-05 + }, + "wall_ms": 433, + "server_latency_ms": null, + "tokens_in": 120, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-15", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.69991, + "p_label": 0.69991, + "distribution": { + "sales": 0.69991, + "technical": 0.291766, + "account": 0.004162, + "billing": 0.004162 + }, + "wall_ms": 435, + "server_latency_ms": null, + "tokens_in": 121, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-16", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.97776, + "p_label": 0.97776, + "distribution": { + "account": 0.97776, + "technical": 0.020293, + "sales": 0.00147, + "billing": 0.000477 + }, + "wall_ms": 433, + "server_latency_ms": null, + "tokens_in": 128, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-17", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.990823, + "p_label": 0.990823, + "distribution": { + "billing": 0.990823, + "sales": 0.006676, + "technical": 0.002456, + "account": 4.5e-05 + }, + "wall_ms": 435, + "server_latency_ms": null, + "tokens_in": 123, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-18", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.999811, + "p_label": 0.999811, + "distribution": { + "technical": 0.999811, + "sales": 0.00014, + "billing": 4e-05, + "account": 9e-06 + }, + "wall_ms": 429, + "server_latency_ms": null, + "tokens_in": 127, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-19", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.793207, + "p_label": 0.793207, + "distribution": { + "sales": 0.793207, + "billing": 0.200554, + "technical": 0.006056, + "account": 0.000183 + }, + "wall_ms": 405, + "server_latency_ms": null, + "tokens_in": 120, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-20", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.998675, + "p_label": 0.998675, + "distribution": { + "account": 0.998675, + "sales": 0.000804, + "technical": 0.000261, + "billing": 0.000261 + }, + "wall_ms": 401, + "server_latency_ms": null, + "tokens_in": 120, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-01", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": false, + "correct": false, + "p_answer": 0.982014, + "p_label": 0.017986, + "distribution": { + "false": 0.982014, + "true": 0.017986 + }, + "wall_ms": 400, + "server_latency_ms": null, + "tokens_in": 102, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-02", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.982014, + "p_label": 0.982014, + "distribution": { + "true": 0.982014, + "false": 0.017986 + }, + "wall_ms": 400, + "server_latency_ms": null, + "tokens_in": 110, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-03", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.992423, + "p_label": 0.992423, + "distribution": { + "false": 0.992423, + "true": 0.007577 + }, + "wall_ms": 400, + "server_latency_ms": null, + "tokens_in": 103, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-04", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": false, + "correct": false, + "p_answer": 0.880797, + "p_label": 0.119203, + "distribution": { + "false": 0.880797, + "true": 0.119203 + }, + "wall_ms": 399, + "server_latency_ms": null, + "tokens_in": 103, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-05", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.989013, + "p_label": 0.989013, + "distribution": { + "false": 0.989013, + "true": 0.010987 + }, + "wall_ms": 397, + "server_latency_ms": null, + "tokens_in": 103, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-06", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.998968, + "p_label": 0.998968, + "distribution": { + "false": 0.998968, + "true": 0.001032 + }, + "wall_ms": 398, + "server_latency_ms": null, + "tokens_in": 106, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-07", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.99593, + "p_label": 0.99593, + "distribution": { + "false": 0.99593, + "true": 0.00407 + }, + "wall_ms": 386, + "server_latency_ms": null, + "tokens_in": 104, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-08", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.993307, + "p_label": 0.993307, + "distribution": { + "false": 0.993307, + "true": 0.006693 + }, + "wall_ms": 388, + "server_latency_ms": null, + "tokens_in": 102, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-09", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.7773, + "p_label": 0.7773, + "distribution": { + "true": 0.7773, + "false": 0.2227 + }, + "wall_ms": 388, + "server_latency_ms": null, + "tokens_in": 102, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-10", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.974043, + "p_label": 0.974043, + "distribution": { + "false": 0.974043, + "true": 0.025957 + }, + "wall_ms": 388, + "server_latency_ms": null, + "tokens_in": 105, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-11", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.984094, + "p_label": 0.984094, + "distribution": { + "false": 0.984094, + "true": 0.015906 + }, + "wall_ms": 387, + "server_latency_ms": null, + "tokens_in": 101, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-12", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.562177, + "p_label": 0.562177, + "distribution": { + "true": 0.562177, + "false": 0.437823 + }, + "wall_ms": 387, + "server_latency_ms": null, + "tokens_in": 108, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-13", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.989013, + "p_label": 0.989013, + "distribution": { + "false": 0.989013, + "true": 0.010987 + }, + "wall_ms": 385, + "server_latency_ms": null, + "tokens_in": 100, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-14", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": false, + "correct": false, + "p_answer": 0.562177, + "p_label": 0.437823, + "distribution": { + "false": 0.562177, + "true": 0.437823 + }, + "wall_ms": 385, + "server_latency_ms": null, + "tokens_in": 101, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-15", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.991423, + "p_label": 0.991423, + "distribution": { + "false": 0.991423, + "true": 0.008577 + }, + "wall_ms": 404, + "server_latency_ms": null, + "tokens_in": 102, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-16", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.914901, + "p_label": 0.914901, + "distribution": { + "true": 0.914901, + "false": 0.085099 + }, + "wall_ms": 403, + "server_latency_ms": null, + "tokens_in": 109, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-17", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.99593, + "p_label": 0.99593, + "distribution": { + "false": 0.99593, + "true": 0.00407 + }, + "wall_ms": 403, + "server_latency_ms": null, + "tokens_in": 104, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-18", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.904651, + "p_label": 0.904651, + "distribution": { + "true": 0.904651, + "false": 0.095349 + }, + "wall_ms": 403, + "server_latency_ms": null, + "tokens_in": 108, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-19", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.984094, + "p_label": 0.984094, + "distribution": { + "false": 0.984094, + "true": 0.015906 + }, + "wall_ms": 402, + "server_latency_ms": null, + "tokens_in": 101, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-20", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.987568, + "p_label": 0.987568, + "distribution": { + "false": 0.987568, + "true": 0.012432 + }, + "wall_ms": 401, + "server_latency_ms": null, + "tokens_in": 101, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-01", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.754915, + "p_label": 0.754915, + "distribution": { + "true": 0.754915, + "false": 0.245085 + }, + "wall_ms": 400, + "server_latency_ms": null, + "tokens_in": 114, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-02", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.731059, + "p_label": 0.731059, + "distribution": { + "true": 0.731059, + "false": 0.268941 + }, + "wall_ms": 399, + "server_latency_ms": null, + "tokens_in": 106, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-03", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.880797, + "p_label": 0.880797, + "distribution": { + "false": 0.880797, + "true": 0.119203 + }, + "wall_ms": 411, + "server_latency_ms": null, + "tokens_in": 106, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-04", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.998675, + "p_label": 0.998675, + "distribution": { + "false": 0.998675, + "true": 0.001325 + }, + "wall_ms": 413, + "server_latency_ms": null, + "tokens_in": 107, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-05", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.679179, + "p_label": 0.679179, + "distribution": { + "true": 0.679179, + "false": 0.320821 + }, + "wall_ms": 411, + "server_latency_ms": null, + "tokens_in": 118, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-06", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.991423, + "p_label": 0.991423, + "distribution": { + "false": 0.991423, + "true": 0.008577 + }, + "wall_ms": 410, + "server_latency_ms": null, + "tokens_in": 112, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-07", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.957912, + "p_label": 0.957912, + "distribution": { + "true": 0.957912, + "false": 0.042088 + }, + "wall_ms": 411, + "server_latency_ms": null, + "tokens_in": 108, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-08", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.99539, + "p_label": 0.99539, + "distribution": { + "false": 0.99539, + "true": 0.00461 + }, + "wall_ms": 411, + "server_latency_ms": null, + "tokens_in": 109, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-09", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.7773, + "p_label": 0.7773, + "distribution": { + "true": 0.7773, + "false": 0.2227 + }, + "wall_ms": 411, + "server_latency_ms": null, + "tokens_in": 107, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-10", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.962673, + "p_label": 0.962673, + "distribution": { + "false": 0.962673, + "true": 0.037327 + }, + "wall_ms": 411, + "server_latency_ms": null, + "tokens_in": 113, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-11", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.592667, + "p_label": 0.592667, + "distribution": { + "true": 0.592667, + "false": 0.407333 + }, + "wall_ms": 405, + "server_latency_ms": null, + "tokens_in": 113, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-12", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.985936, + "p_label": 0.985936, + "distribution": { + "false": 0.985936, + "true": 0.014064 + }, + "wall_ms": 407, + "server_latency_ms": null, + "tokens_in": 115, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-13", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.867036, + "p_label": 0.867036, + "distribution": { + "true": 0.867036, + "false": 0.132964 + }, + "wall_ms": 407, + "server_latency_ms": null, + "tokens_in": 111, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-14", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.991423, + "p_label": 0.991423, + "distribution": { + "false": 0.991423, + "true": 0.008577 + }, + "wall_ms": 406, + "server_latency_ms": null, + "tokens_in": 112, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-15", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": false, + "correct": false, + "p_answer": 0.705785, + "p_label": 0.294215, + "distribution": { + "false": 0.705785, + "true": 0.294215 + }, + "wall_ms": 406, + "server_latency_ms": null, + "tokens_in": 110, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-16", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.851953, + "p_label": 0.851953, + "distribution": { + "false": 0.851953, + "true": 0.148047 + }, + "wall_ms": 406, + "server_latency_ms": null, + "tokens_in": 119, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-17", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.531209, + "p_label": 0.531209, + "distribution": { + "true": 0.531209, + "false": 0.468791 + }, + "wall_ms": 406, + "server_latency_ms": null, + "tokens_in": 108, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-18", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.924142, + "p_label": 0.924142, + "distribution": { + "false": 0.924142, + "true": 0.075858 + }, + "wall_ms": 404, + "server_latency_ms": null, + "tokens_in": 108, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-19", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.679179, + "p_label": 0.679179, + "distribution": { + "true": 0.679179, + "false": 0.320821 + }, + "wall_ms": 363, + "server_latency_ms": null, + "tokens_in": 106, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-20", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.993307, + "p_label": 0.993307, + "distribution": { + "false": 0.993307, + "true": 0.006693 + }, + "wall_ms": 361, + "server_latency_ms": null, + "tokens_in": 113, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-01", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.957912, + "p_label": 0.957912, + "distribution": { + "true": 0.957912, + "false": 0.042088 + }, + "wall_ms": 361, + "server_latency_ms": null, + "tokens_in": 73, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-02", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.999512, + "p_label": 0.999512, + "distribution": { + "false": 0.999512, + "true": 0.000488 + }, + "wall_ms": 361, + "server_latency_ms": null, + "tokens_in": 74, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-03", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.99478, + "p_label": 0.99478, + "distribution": { + "true": 0.99478, + "false": 0.00522 + }, + "wall_ms": 361, + "server_latency_ms": null, + "tokens_in": 64, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-04", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.987568, + "p_label": 0.987568, + "distribution": { + "true": 0.987568, + "false": 0.012432 + }, + "wall_ms": 359, + "server_latency_ms": null, + "tokens_in": 66, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-05", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.592667, + "p_label": 0.592667, + "distribution": { + "true": 0.592667, + "false": 0.407333 + }, + "wall_ms": 359, + "server_latency_ms": null, + "tokens_in": 79, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-06", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.999196, + "p_label": 0.999196, + "distribution": { + "false": 0.999196, + "true": 0.000804 + }, + "wall_ms": 359, + "server_latency_ms": null, + "tokens_in": 73, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-07", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.997527, + "p_label": 0.997527, + "distribution": { + "false": 0.997527, + "true": 0.002473 + }, + "wall_ms": 369, + "server_latency_ms": null, + "tokens_in": 70, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-08", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.997199, + "p_label": 0.997199, + "distribution": { + "true": 0.997199, + "false": 0.002801 + }, + "wall_ms": 375, + "server_latency_ms": null, + "tokens_in": 73, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-09", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.99593, + "p_label": 0.99593, + "distribution": { + "true": 0.99593, + "false": 0.00407 + }, + "wall_ms": 375, + "server_latency_ms": null, + "tokens_in": 66, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-10", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.904651, + "p_label": 0.904651, + "distribution": { + "false": 0.904651, + "true": 0.095349 + }, + "wall_ms": 375, + "server_latency_ms": null, + "tokens_in": 70, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-11", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.957912, + "p_label": 0.957912, + "distribution": { + "true": 0.957912, + "false": 0.042088 + }, + "wall_ms": 374, + "server_latency_ms": null, + "tokens_in": 78, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-12", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.977023, + "p_label": 0.977023, + "distribution": { + "true": 0.977023, + "false": 0.022977 + }, + "wall_ms": 374, + "server_latency_ms": null, + "tokens_in": 70, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-13", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.998968, + "p_label": 0.998968, + "distribution": { + "false": 0.998968, + "true": 0.001032 + }, + "wall_ms": 372, + "server_latency_ms": null, + "tokens_in": 67, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-14", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.705785, + "p_label": 0.705785, + "distribution": { + "true": 0.705785, + "false": 0.294215 + }, + "wall_ms": 371, + "server_latency_ms": null, + "tokens_in": 69, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-15", + "set": "fact", + "kind": "noul", + "label": false, + "answer": true, + "correct": false, + "p_answer": 0.562177, + "p_label": 0.437823, + "distribution": { + "true": 0.562177, + "false": 0.437823 + }, + "wall_ms": 317, + "server_latency_ms": null, + "tokens_in": 67, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-16", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.651355, + "p_label": 0.651355, + "distribution": { + "true": 0.651355, + "false": 0.348645 + }, + "wall_ms": 317, + "server_latency_ms": null, + "tokens_in": 76, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-17", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.999569, + "p_label": 0.999569, + "distribution": { + "false": 0.999569, + "true": 0.000431 + }, + "wall_ms": 317, + "server_latency_ms": null, + "tokens_in": 68, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-18", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.592667, + "p_label": 0.592667, + "distribution": { + "true": 0.592667, + "false": 0.407333 + }, + "wall_ms": 316, + "server_latency_ms": null, + "tokens_in": 77, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-19", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.998073, + "p_label": 0.998073, + "distribution": { + "false": 0.998073, + "true": 0.001927 + }, + "wall_ms": 315, + "server_latency_ms": null, + "tokens_in": 65, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-20", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.982014, + "p_label": 0.982014, + "distribution": { + "true": 0.982014, + "false": 0.017986 + }, + "wall_ms": 314, + "server_latency_ms": null, + "tokens_in": 67, + "tokens_out": 2, + "error": null + } + ] + }, + "notes": [ + "Measured by scripts/ainode-bench.py decide in 7s over 110 labeled items from items.json (decide-110); nothing was loaded, unloaded or restarted.", + "Accuracy is the weakest number here: the failure mode automation cares about is a wrong answer at high confidence, which is what the Brier score, the calibration error and the wrong-at-threshold counts measure.", + "Confidence is the probability the backend put on the answer it gave. Brier is a single term on the labeled option's probability, and the calibration error uses 5 bins over the reliability table in the same block.", + "Cost is $0 for the chat backend: it runs on our own hardware and nobody bills per token for it. The electricity is real and is not a number this record claims to have measured.", + "The chat backend is the fallback instrument: lettered options with thinking off, and a distribution softmaxed from the top logprobs of the one letter token. Those probabilities are over letters rather than over meanings, so they are a weaker statement than a decision endpoint's." + ], + "source": "scripts/ainode-bench.py decide" +} diff --git a/bench/results/20261002-005750-nvidia-nemotron-3-nano-4b-bf16-spark-4-stacked-beside-jeb-vllm-0_27_1-bf16-agentic.json b/bench/results/20261002-005750-nvidia-nemotron-3-nano-4b-bf16-spark-4-stacked-beside-jeb-vllm-0_27_1-bf16-agentic.json new file mode 100644 index 00000000..ea1b787f --- /dev/null +++ b/bench/results/20261002-005750-nvidia-nemotron-3-nano-4b-bf16-spark-4-stacked-beside-jeb-vllm-0_27_1-bf16-agentic.json @@ -0,0 +1,328 @@ +{ + "schema": 1, + "stamp": "20261002-005750", + "label": "spark-4 stacked beside jeb, vllm 0.27.1 bf16", + "model": { + "id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16", + "name": "NVIDIA Nemotron 3 Nano 4B BF16", + "params_b": 4.0, + "vision": false, + "arch": "dense", + "active_b": 4.0, + "quant": "BF16" + }, + "placement": { + "engine": "vllm", + "node": "Spark-4-GX10", + "gpu": "NVIDIA GB10", + "ainode": "0.5.33", + "tp": 1, + "gpus": 1, + "stacked_with": [ + "frontier-infra/jebadiah-9b-v2", + "frontier-infra/jebadiah-4b-v2" + ], + "port": 8000 + }, + "settings": { + "groups": [ + "A", + "B", + "C", + "D", + "E", + "F", + "G" + ], + "needle_tokens": [ + 8000, + 48000 + ], + "quick": false, + "temperature": 1.0, + "timeout_s": 900, + "thinking_switch": "enable_thinking", + "probes_requested": 23 + }, + "agentic": { + "score": { + "pass": 17, + "total": 23 + }, + "groups": { + "A": { + "pass": 3, + "total": 4 + }, + "B": { + "pass": 3, + "total": 4 + }, + "C": { + "pass": 1, + "total": 3 + }, + "D": { + "pass": 3, + "total": 4 + }, + "E": { + "pass": 2, + "total": 2 + }, + "F": { + "pass": 1, + "total": 1 + }, + "G": { + "pass": 4, + "total": 5 + } + }, + "endpoint": "http://100.72.9.84:3000/v1", + "protocol": { + "groups": [ + "A", + "B", + "C", + "D", + "E", + "F", + "G" + ], + "needle_tokens": [ + 8000, + 48000 + ], + "temperature": 1.0, + "timeout_s": 900, + "thinking_switch": "enable_thinking", + "max_turns": 4, + "verdicts": "mechanical; group C is executed, group G is a judged tool trace" + }, + "probes": [ + { + "id": "A1_format", + "group": "A", + "pass": true, + "wall_s": 20.53, + "completion_tokens": 455, + "note": "5 lines, names in caps, numbering loose", + "excerpt": "N. JUPITER N. SATURN N. URANUS N. NEPTUNE N. EARTH" + }, + { + "id": "A2_json_only", + "group": "A", + "pass": false, + "wall_s": 5.82, + "completion_tokens": 128, + "note": "landlocked is False, Austria is landlocked", + "excerpt": "{\"city\":\"Vienna\",\"population_millions\":0.2,\"landlocked\":false}" + }, + { + "id": "A3_constraints", + "group": "A", + "pass": true, + "wall_s": 26.2, + "completion_tokens": 581, + "note": "no z, no 'the', 3 full stops", + "excerpt": "Bicycle pedals rotate wheel. Chain transfers power to rear sprocket. Rider pushes forward gliding smoothly." + }, + { + "id": "A4_persona", + "group": "A", + "pass": true, + "wall_s": 41.74, + "completion_tokens": 926, + "note": "stayed in character", + "excerpt": "Ahoy Captain! A mutex is a mutex that keeps two pirates from both grabbin' the same treasure at once. It's a proper way to share loot without meanness. Arr!" + }, + { + "id": "B1_tool_single", + "group": "B", + "pass": true, + "wall_s": 3.56, + "completion_tokens": 76, + "note": "get_weather({\"city\": \"Tokyo\", \"unit\": \"c\"})", + "excerpt": "" + }, + { + "id": "B2_tool_parallel", + "group": "B", + "pass": false, + "wall_s": 5.92, + "completion_tokens": 127, + "note": "2 call(s), wanted 3: get_weather({\"city\": \"Paris\", \"unit\": \"c\"}), get_weather({\"city\": \"Cairo\", \"unit\": \"c\"})", + "excerpt": "" + }, + { + "id": "B3_tool_not_needed", + "group": "B", + "pass": true, + "wall_s": 21.37, + "completion_tokens": 473, + "note": "answered without a tool", + "excerpt": "Crimson leaves descend Wind whispers soft, good\u2011bye here Night glows amber light" + }, + { + "id": "B4_tool_roundtrip", + "group": "B", + "pass": true, + "wall_s": 5.85, + "completion_tokens": 122, + "note": "used the tool result", + "excerpt": "The current weather in Tokyo is 31\u00b0C with thunderstorms." + }, + { + "id": "C1_ttl_cache", + "group": "C", + "pass": false, + "wall_s": 55.46, + "completion_tokens": 1232, + "note": "AssertionError", + "excerpt": "```python import time class TTLCache: def __init__(self, capacity: int, ttl: float, clock: callable = time.time): self.capacity = capacity self.ttl = ttl self.clock = clock self._cache = {} # key -> (value, timestamp) self._access_counter = {} # key -> int (access order) self._next_access = 0 def _m..." + }, + { + "id": "C2_intervals", + "group": "C", + "pass": false, + "wall_s": 39.44, + "completion_tokens": 875, + "note": "AssertionError", + "excerpt": "```python def merge_intervals(intervals): \"\"\"Merge overlapping or touching [start, end] intervals and return a sorted list.\"\"\" if not intervals: return [] # sort by start sorted_intervals = sorted(intervals, key=lambda x: x[0]) merged = [] cur_start, cur_end = sorted_intervals[0] for s, e in sorted_..." + }, + { + "id": "C3_bugfix", + "group": "C", + "pass": true, + "wall_s": 16.16, + "completion_tokens": 358, + "note": "hidden asserts passed", + "excerpt": "```python import re from collections import Counter def top_k_words(text, k): words = [w.lower() for w in re.findall(r'[a-zA-Z]+', text)] counts = Counter(words) items = sorted(counts.items(), key=lambda kv: (-kv[1], kv[0])) return [w for w, _ in items[:k]] ```" + }, + { + "id": "D1_decimal", + "group": "D", + "pass": true, + "wall_s": 3.09, + "completion_tokens": 65, + "note": "said 9.9", + "excerpt": "9.9" + }, + { + "id": "D2_strawberry", + "group": "D", + "pass": true, + "wall_s": 5.58, + "completion_tokens": 121, + "note": "said 3", + "excerpt": "3" + }, + { + "id": "D3_sisters", + "group": "D", + "pass": true, + "wall_s": 17.95, + "completion_tokens": 393, + "note": "said 2", + "excerpt": "2" + }, + { + "id": "D4_knights", + "group": "D", + "pass": false, + "wall_s": 20.26, + "completion_tokens": 447, + "note": "wanted A=Knight, B=Knave", + "excerpt": "A=Knight, B=Knave." + }, + { + "id": "E_needle_8000", + "group": "E", + "pass": true, + "wall_s": 3.43, + "completion_tokens": 35, + "note": "found the password; prompt_tokens=6516", + "excerpt": "violet-otter-2931" + }, + { + "id": "E_needle_48000", + "group": "E", + "pass": true, + "wall_s": 15.18, + "completion_tokens": 57, + "note": "found the password; prompt_tokens=38929", + "excerpt": "violet-otter-2931" + }, + { + "id": "F1_thinking_off", + "group": "F", + "pass": true, + "wall_s": 0.23, + "completion_tokens": 4, + "note": "answered with no reasoning channel", + "excerpt": "391" + }, + { + "id": "G1_tool_chain", + "group": "G", + "pass": true, + "wall_s": 5.0, + "completion_tokens": 101, + "note": "list_files -> read_file; answered 137", + "excerpt": "137" + }, + { + "id": "G2_tool_error", + "group": "G", + "pass": true, + "wall_s": 8.84, + "completion_tokens": 185, + "note": "called the tool with 'San Francisco' after the error", + "excerpt": "The current temperature in San Francisco is **18 \u00b0C**." + }, + { + "id": "G3_arg_schema", + "group": "G", + "pass": true, + "wall_s": 7.32, + "completion_tokens": 160, + "note": "mode='daily' retries=3 notify=True", + "excerpt": "[[\"schedule_job\", \"{\\\"mode\\\": \\\"daily\\\", \\\"retries\\\": 3, \\\"notify\\\": true}\"]]" + }, + { + "id": "G4_structured", + "group": "G", + "pass": true, + "wall_s": 7.08, + "completion_tokens": 135, + "note": "four keys, right types, severity=2 [json_schema]", + "excerpt": "{ \"title\": \"Login service outage due to bad deploy\", \"severity\": 2, \"resolved\": true, \"tags\": [\"login\", \"deploy\"] }" + }, + { + "id": "G5_persistence", + "group": "G", + "pass": false, + "wall_s": 40.33, + "completion_tokens": 881, + "note": "3/4 turns ended with DONE; missed turn(s) [4]", + "excerpt": "" + } + ], + "needle": { + "8000": true, + "48000": true + }, + "thinking_off_supported": true, + "vision_supported": null, + "structured_output_mode": "json_schema" + }, + "notes": [ + "Measured by scripts/ainode-bench.py agentic in 376s; nothing loaded, unloaded or restarted.", + "Every verdict is mechanical: no judge model and no reply was read by a person to reach a score.", + "Group C runs the model's own code against asserts it never saw, in a subprocess on the machine driving the bench.", + "Group G drives real tool loops and judges the call trace: the order of the calls, the arguments, and whether a stated fact came back from a tool.", + "Groups run: A, B, C, D, E, F, G. A group that did not run is absent from the score rather than counted as zero." + ], + "source": "scripts/ainode-bench.py agentic" +} diff --git a/bench/results/20261002-010139-nvidia-nemotron-3-nano-4b-bf16-spark-4-stacked-beside-jeb-4b-and-9b-vllm-0_27_1-bf16.json b/bench/results/20261002-010139-nvidia-nemotron-3-nano-4b-bf16-spark-4-stacked-beside-jeb-4b-and-9b-vllm-0_27_1-bf16.json new file mode 100644 index 00000000..9bd256d7 --- /dev/null +++ b/bench/results/20261002-010139-nvidia-nemotron-3-nano-4b-bf16-spark-4-stacked-beside-jeb-4b-and-9b-vllm-0_27_1-bf16.json @@ -0,0 +1,155 @@ +{ + "schema": 1, + "stamp": "20261002-010139", + "label": "spark-4 stacked beside jeb 4b and 9b, vllm 0.27.1 bf16", + "model": { + "id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16", + "name": "NVIDIA Nemotron 3 Nano 4B BF16", + "params_b": 4.0, + "vision": false, + "arch": "dense", + "active_b": 4.0, + "quant": "BF16" + }, + "placement": { + "engine": "vllm", + "node": "Spark-4-GX10", + "gpu": "NVIDIA GB10", + "ainode": "0.5.33", + "tp": 1, + "gpus": 1, + "stacked_with": [ + "frontier-infra/jebadiah-9b-v2", + "frontier-infra/jebadiah-4b-v2" + ] + }, + "settings": { + "thinking": false, + "max_tokens": 200, + "temperature": null, + "sustained_tokens": 1500, + "reasoning_tokens": 600, + "depths_requested": [ + 4000, + 16000, + 32000, + 60000 + ], + "streams_requested": [ + 1, + 2, + 4, + 8, + 16 + ], + "chars_per_token_probe": 5.38 + }, + "results": { + "single_stream": { + "decode_tok_s": 22.2, + "gen_tokens": 172, + "ttft_ms": 124, + "prompt_tokens": 81 + }, + "prefill": [ + { + "ttft_ms": 1186, + "prompt_tokens": 3965, + "prefill_tok_s": 3344.0, + "decode_tok_s": 22.26 + }, + { + "ttft_ms": 4732, + "prompt_tokens": 15842, + "prefill_tok_s": 3347.9, + "decode_tok_s": 21.95 + }, + { + "ttft_ms": 10272, + "prompt_tokens": 31612, + "prefill_tok_s": 3077.6, + "decode_tok_s": 21.61 + }, + { + "ttft_ms": 21658, + "prompt_tokens": 59188, + "prefill_tok_s": 2732.8, + "decode_tok_s": 20.98 + } + ], + "sustained": { + "decode_tok_s": 22.25, + "gen_tokens": 1208, + "wall_s": 54.383, + "ttft_ms": 129 + }, + "concurrency": [ + { + "streams": 1, + "aggregate_tok_s": 21.6, + "ok": 1, + "failed": 0, + "wall_s": 7.71, + "per_stream_tok_s": 22.3, + "median_ttft_ms": 259 + }, + { + "streams": 2, + "aggregate_tok_s": 41.5, + "ok": 2, + "failed": 0, + "wall_s": 8.59, + "per_stream_tok_s": 22.7, + "median_ttft_ms": 362 + }, + { + "streams": 4, + "aggregate_tok_s": 81.0, + "ok": 4, + "failed": 0, + "wall_s": 9.49, + "per_stream_tok_s": 22.0, + "median_ttft_ms": 440 + }, + { + "streams": 8, + "aggregate_tok_s": 153.2, + "ok": 8, + "failed": 0, + "wall_s": 9.91, + "per_stream_tok_s": 20.8, + "median_ttft_ms": 337 + }, + { + "streams": 16, + "aggregate_tok_s": 254.2, + "ok": 16, + "failed": 0, + "wall_s": 11.24, + "per_stream_tok_s": 18.3, + "median_ttft_ms": 526 + } + ], + "reasoning_tax": { + "thinking_off_wall_s": 14.849, + "thinking_off_tokens": 328, + "thinking_on_wall_s": 18.147, + "thinking_on_tokens": 399, + "tax_x": 1.22 + }, + "telemetry": { + "gpu_mem_used_gb": 75.4, + "gpu_mem_total_gb": 121.6, + "temp_c": 51, + "samples": 101, + "reading": "peak observed over the run; sampled from /api/nodes", + "temp_c_at_start": 51 + } + }, + "notes": [ + "Measured by scripts/ainode-bench.py in 217s; nothing loaded, unloaded or restarted.", + "prefill_tok_s is prompt_tokens/TTFT, a floor: TTFT includes queueing, and the OpenAI API exposes no internal prefill timing.", + "Sections other than reasoning_tax ran with enable_thinking=false." + ], + "source": "scripts/ainode-bench.py" +} diff --git a/bench/results/20261002-010541-nemotron-nano-4b-q4km-v100-pollux-one-v100-32gb-llama_cpp-server-cuda-official-q4_k_m-gguf-outside-ainode-decide.json b/bench/results/20261002-010541-nemotron-nano-4b-q4km-v100-pollux-one-v100-32gb-llama_cpp-server-cuda-official-q4_k_m-gguf-outside-ainode-decide.json new file mode 100644 index 00000000..b311b25c --- /dev/null +++ b/bench/results/20261002-010541-nemotron-nano-4b-q4km-v100-pollux-one-v100-32gb-llama_cpp-server-cuda-official-q4_k_m-gguf-outside-ainode-decide.json @@ -0,0 +1,2657 @@ +{ + "schema": 1, + "stamp": "20261002-010541", + "label": "pollux one v100 32gb, llama.cpp server-cuda, official q4_k_m gguf, outside ainode", + "model": { + "id": "nemotron-nano-4b-q4km-v100" + }, + "placement": {}, + "settings": { + "backend": "chat", + "endpoint": "http://100.124.227.93:8101/v1/chat/completions", + "model_requested": "nemotron-nano-4b-q4km-v100", + "items_file": "bench/decide/items.json", + "item_set": "decide-110", + "items": 110, + "sets": [ + "route", + "triage", + "urgency", + "pr_safe", + "fact" + ], + "concurrency": 8, + "timeout_s": 120 + }, + "decide": { + "backend": "chat", + "endpoint": "http://100.124.227.93:8101/v1/chat/completions", + "model_reported": "nemotron-nano-4b-q4km-v100", + "item_set": { + "id": "decide-110", + "version": 1, + "file": "items.json", + "count": 110, + "sets": { + "route": 30, + "triage": 20, + "urgency": 20, + "pr_safe": 20, + "fact": 20 + } + }, + "protocol": { + "backend": "chat", + "endpoint": "http://100.124.227.93:8101/v1/chat/completions", + "timeout_s": 120, + "input_usd_per_mtok": 0.0, + "output_usd_per_mtok": 0.0, + "concurrency": 8, + "bins": 5, + "thresholds": [ + "0.8", + "0.9" + ], + "confidence": "the probability the backend put on the answer it gave; its reported confidence when it gave no distribution", + "brier": "one term, on the labeled option's probability" + }, + "overall": { + "n": 110, + "answered": 110, + "errors": 0, + "accuracy": 0.918, + "brier": 0.065, + "ece": 0.023, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 4, + "accuracy": 0.5, + "confidence": 0.562 + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 18, + "accuracy": 0.778, + "confidence": 0.676 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 88, + "accuracy": 0.966, + "confidence": 0.971 + } + ], + "thresholds": { + "0.8": { + "kept": 88, + "wrong": 3, + "abstained": 22, + "no_confidence": 0 + }, + "0.9": { + "kept": 77, + "wrong": 2, + "abstained": 33, + "no_confidence": 0 + } + }, + "tokens": { + "in": 12151, + "out": 220 + }, + "cost_usd": 0.0, + "p50_ms": 4013, + "p95_ms": 34128 + }, + "sets": { + "route": { + "n": 30, + "answered": 30, + "errors": 0, + "accuracy": 0.933, + "brier": 0.053, + "ece": 0.017, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 4, + "accuracy": 0.75, + "confidence": 0.672 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 26, + "accuracy": 0.962, + "confidence": 0.969 + } + ], + "thresholds": { + "0.8": { + "kept": 26, + "wrong": 1, + "abstained": 4, + "no_confidence": 0 + }, + "0.9": { + "kept": 21, + "wrong": 0, + "abstained": 9, + "no_confidence": 0 + } + }, + "tokens": { + "in": 3914, + "out": 60 + }, + "cost_usd": 0.0, + "p50_ms": 3949, + "p95_ms": 36081 + }, + "triage": { + "n": 20, + "answered": 20, + "errors": 0, + "accuracy": 1.0, + "brier": 0.008, + "ece": 0.044, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 2, + "accuracy": 1.0, + "confidence": 0.755 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 18, + "accuracy": 1.0, + "confidence": 0.978 + } + ], + "thresholds": { + "0.8": { + "kept": 18, + "wrong": 0, + "abstained": 2, + "no_confidence": 0 + }, + "0.9": { + "kept": 16, + "wrong": 0, + "abstained": 4, + "no_confidence": 0 + } + }, + "tokens": { + "in": 2475, + "out": 40 + }, + "cost_usd": 0.0, + "p50_ms": 3946, + "p95_ms": 4080 + }, + "urgency": { + "n": 20, + "answered": 20, + "errors": 0, + "accuracy": 0.8, + "brier": 0.133, + "ece": 0.124, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 1, + "accuracy": 0.0, + "confidence": 0.553 + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 2, + "accuracy": 0.5, + "confidence": 0.673 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 17, + "accuracy": 0.882, + "confidence": 0.975 + } + ], + "thresholds": { + "0.8": { + "kept": 17, + "wrong": 2, + "abstained": 3, + "no_confidence": 0 + }, + "0.9": { + "kept": 16, + "wrong": 2, + "abstained": 4, + "no_confidence": 0 + } + }, + "tokens": { + "in": 2095, + "out": 40 + }, + "cost_usd": 0.0, + "p50_ms": 4143, + "p95_ms": 4167 + }, + "pr_safe": { + "n": 20, + "answered": 20, + "errors": 0, + "accuracy": 0.9, + "brier": 0.086, + "ece": 0.084, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 2, + "accuracy": 0.5, + "confidence": 0.555 + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 6, + "accuracy": 0.833, + "confidence": 0.678 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 12, + "accuracy": 1.0, + "confidence": 0.946 + } + ], + "thresholds": { + "0.8": { + "kept": 12, + "wrong": 0, + "abstained": 8, + "no_confidence": 0 + }, + "0.9": { + "kept": 9, + "wrong": 0, + "abstained": 11, + "no_confidence": 0 + } + }, + "tokens": { + "in": 2235, + "out": 40 + }, + "cost_usd": 0.0, + "p50_ms": 4042, + "p95_ms": 4087 + }, + "fact": { + "n": 20, + "answered": 20, + "errors": 0, + "accuracy": 0.95, + "brier": 0.049, + "ece": 0.057, + "bins": [ + { + "lo": 0.0, + "hi": 0.2, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.2, + "hi": 0.4, + "count": 0, + "accuracy": null, + "confidence": null + }, + { + "lo": 0.4, + "hi": 0.6, + "count": 1, + "accuracy": 1.0, + "confidence": 0.585 + }, + { + "lo": 0.6, + "hi": 0.8, + "count": 4, + "accuracy": 0.75, + "confidence": 0.638 + }, + { + "lo": 0.8, + "hi": 1.0, + "count": 15, + "accuracy": 1.0, + "confidence": 0.982 + } + ], + "thresholds": { + "0.8": { + "kept": 15, + "wrong": 0, + "abstained": 5, + "no_confidence": 0 + }, + "0.9": { + "kept": 15, + "wrong": 0, + "abstained": 5, + "no_confidence": 0 + } + }, + "tokens": { + "in": 1432, + "out": 40 + }, + "cost_usd": 0.0, + "p50_ms": 3993, + "p95_ms": 4083 + } + }, + "rows": [ + { + "id": "route-01", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.98309, + "p_label": 0.98309, + "distribution": { + "code": 0.98309, + "chat": 0.015006, + "vision": 0.001772, + "long_document": 0.000132 + }, + "wall_ms": 34125, + "server_latency_ms": null, + "tokens_in": 131, + "tokens_out": 2, + "error": null + }, + { + "id": "route-02", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.998859, + "p_label": 0.998859, + "distribution": { + "chat": 0.998859, + "vision": 0.000766, + "code": 0.000322, + "long_document": 5.4e-05 + }, + "wall_ms": 34128, + "server_latency_ms": null, + "tokens_in": 129, + "tokens_out": 2, + "error": null + }, + { + "id": "route-03", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.998124, + "p_label": 0.998124, + "distribution": { + "vision": 0.998124, + "chat": 0.001212, + "code": 0.000507, + "long_document": 0.000158 + }, + "wall_ms": 34123, + "server_latency_ms": null, + "tokens_in": 127, + "tokens_out": 2, + "error": null + }, + { + "id": "route-04", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.978918, + "p_label": 0.978918, + "distribution": { + "long_document": 0.978918, + "chat": 0.013846, + "vision": 0.005892, + "code": 0.001344 + }, + "wall_ms": 36085, + "server_latency_ms": null, + "tokens_in": 137, + "tokens_out": 2, + "error": null + }, + { + "id": "route-05", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.850772, + "p_label": 0.850772, + "distribution": { + "code": 0.850772, + "chat": 0.143356, + "vision": 0.004544, + "long_document": 0.001327 + }, + "wall_ms": 34129, + "server_latency_ms": null, + "tokens_in": 135, + "tokens_out": 2, + "error": null + }, + { + "id": "route-06", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.999426, + "p_label": 0.999426, + "distribution": { + "chat": 0.999426, + "vision": 0.000339, + "code": 0.000219, + "long_document": 1.6e-05 + }, + "wall_ms": 36081, + "server_latency_ms": null, + "tokens_in": 128, + "tokens_out": 2, + "error": null + }, + { + "id": "route-07", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.989763, + "p_label": 0.989763, + "distribution": { + "vision": 0.989763, + "chat": 0.005925, + "code": 0.00362, + "long_document": 0.000692 + }, + "wall_ms": 36079, + "server_latency_ms": null, + "tokens_in": 131, + "tokens_out": 2, + "error": null + }, + { + "id": "route-08", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.997697, + "p_label": 0.997697, + "distribution": { + "long_document": 0.997697, + "chat": 0.001849, + "code": 0.000281, + "vision": 0.000173 + }, + "wall_ms": 36075, + "server_latency_ms": null, + "tokens_in": 136, + "tokens_out": 2, + "error": null + }, + { + "id": "route-09", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.898849, + "p_label": 0.898849, + "distribution": { + "code": 0.898849, + "chat": 0.098749, + "vision": 0.002209, + "long_document": 0.000193 + }, + "wall_ms": 3908, + "server_latency_ms": null, + "tokens_in": 130, + "tokens_out": 2, + "error": null + }, + { + "id": "route-10", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.999359, + "p_label": 0.999359, + "distribution": { + "chat": 0.999359, + "vision": 0.000408, + "code": 0.000216, + "long_document": 1.7e-05 + }, + "wall_ms": 3909, + "server_latency_ms": null, + "tokens_in": 128, + "tokens_out": 2, + "error": null + }, + { + "id": "route-11", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.895357, + "p_label": 0.895357, + "distribution": { + "vision": 0.895357, + "chat": 0.097466, + "code": 0.004713, + "long_document": 0.002464 + }, + "wall_ms": 3902, + "server_latency_ms": null, + "tokens_in": 131, + "tokens_out": 2, + "error": null + }, + { + "id": "route-12", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.993373, + "p_label": 0.993373, + "distribution": { + "long_document": 0.993373, + "code": 0.003999, + "chat": 0.002141, + "vision": 0.000488 + }, + "wall_ms": 3897, + "server_latency_ms": null, + "tokens_in": 137, + "tokens_out": 2, + "error": null + }, + { + "id": "route-13", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.622577, + "p_label": 0.622577, + "distribution": { + "code": 0.622577, + "chat": 0.374172, + "vision": 0.002857, + "long_document": 0.000394 + }, + "wall_ms": 3964, + "server_latency_ms": null, + "tokens_in": 132, + "tokens_out": 2, + "error": null + }, + { + "id": "route-14", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.995489, + "p_label": 0.995489, + "distribution": { + "chat": 0.995489, + "code": 0.003587, + "vision": 0.000853, + "long_document": 7.1e-05 + }, + "wall_ms": 3959, + "server_latency_ms": null, + "tokens_in": 124, + "tokens_out": 2, + "error": null + }, + { + "id": "route-15", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.976535, + "p_label": 0.976535, + "distribution": { + "vision": 0.976535, + "chat": 0.016161, + "code": 0.005882, + "long_document": 0.001422 + }, + "wall_ms": 3954, + "server_latency_ms": null, + "tokens_in": 126, + "tokens_out": 2, + "error": null + }, + { + "id": "route-16", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.658995, + "p_label": 0.658995, + "distribution": { + "long_document": 0.658995, + "chat": 0.138546, + "code": 0.115066, + "vision": 0.087393 + }, + "wall_ms": 3949, + "server_latency_ms": null, + "tokens_in": 129, + "tokens_out": 2, + "error": null + }, + { + "id": "route-17", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "chat", + "correct": false, + "p_answer": 0.889103, + "p_label": 0.109226, + "distribution": { + "chat": 0.889103, + "code": 0.109226, + "vision": 0.001271, + "long_document": 0.0004 + }, + "wall_ms": 3998, + "server_latency_ms": null, + "tokens_in": 130, + "tokens_out": 2, + "error": null + }, + { + "id": "route-18", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.996778, + "p_label": 0.996778, + "distribution": { + "chat": 0.996778, + "vision": 0.002004, + "code": 0.001131, + "long_document": 8.7e-05 + }, + "wall_ms": 3993, + "server_latency_ms": null, + "tokens_in": 129, + "tokens_out": 2, + "error": null + }, + { + "id": "route-19", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.983465, + "p_label": 0.983465, + "distribution": { + "vision": 0.983465, + "code": 0.012532, + "chat": 0.003061, + "long_document": 0.000942 + }, + "wall_ms": 3988, + "server_latency_ms": null, + "tokens_in": 128, + "tokens_out": 2, + "error": null + }, + { + "id": "route-20", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.960141, + "p_label": 0.960141, + "distribution": { + "long_document": 0.960141, + "chat": 0.033618, + "code": 0.004723, + "vision": 0.001517 + }, + "wall_ms": 3983, + "server_latency_ms": null, + "tokens_in": 135, + "tokens_out": 2, + "error": null + }, + { + "id": "route-21", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.975568, + "p_label": 0.975568, + "distribution": { + "code": 0.975568, + "chat": 0.021695, + "vision": 0.002527, + "long_document": 0.000209 + }, + "wall_ms": 3945, + "server_latency_ms": null, + "tokens_in": 125, + "tokens_out": 2, + "error": null + }, + { + "id": "route-22", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.998908, + "p_label": 0.998908, + "distribution": { + "chat": 0.998908, + "vision": 0.00062, + "code": 0.000444, + "long_document": 2.8e-05 + }, + "wall_ms": 3940, + "server_latency_ms": null, + "tokens_in": 127, + "tokens_out": 2, + "error": null + }, + { + "id": "route-23", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.974372, + "p_label": 0.974372, + "distribution": { + "vision": 0.974372, + "chat": 0.022008, + "code": 0.002717, + "long_document": 0.000903 + }, + "wall_ms": 3935, + "server_latency_ms": null, + "tokens_in": 130, + "tokens_out": 2, + "error": null + }, + { + "id": "route-24", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.995574, + "p_label": 0.995574, + "distribution": { + "long_document": 0.995574, + "vision": 0.001914, + "chat": 0.001807, + "code": 0.000705 + }, + "wall_ms": 3930, + "server_latency_ms": null, + "tokens_in": 138, + "tokens_out": 2, + "error": null + }, + { + "id": "route-25", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "chat", + "correct": false, + "p_answer": 0.638192, + "p_label": 0.3602, + "distribution": { + "chat": 0.638192, + "code": 0.3602, + "vision": 0.001449, + "long_document": 0.000159 + }, + "wall_ms": 3912, + "server_latency_ms": null, + "tokens_in": 135, + "tokens_out": 2, + "error": null + }, + { + "id": "route-26", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.99738, + "p_label": 0.99738, + "distribution": { + "chat": 0.99738, + "vision": 0.001882, + "code": 0.000615, + "long_document": 0.000124 + }, + "wall_ms": 3907, + "server_latency_ms": null, + "tokens_in": 127, + "tokens_out": 2, + "error": null + }, + { + "id": "route-27", + "set": "route", + "kind": "choice", + "label": "vision", + "answer": "vision", + "correct": true, + "p_answer": 0.990029, + "p_label": 0.990029, + "distribution": { + "vision": 0.990029, + "chat": 0.008209, + "code": 0.001362, + "long_document": 0.0004 + }, + "wall_ms": 3901, + "server_latency_ms": null, + "tokens_in": 129, + "tokens_out": 2, + "error": null + }, + { + "id": "route-28", + "set": "route", + "kind": "choice", + "label": "long_document", + "answer": "long_document", + "correct": true, + "p_answer": 0.768426, + "p_label": 0.768426, + "distribution": { + "long_document": 0.768426, + "chat": 0.098051, + "vision": 0.094418, + "code": 0.039105 + }, + "wall_ms": 3897, + "server_latency_ms": null, + "tokens_in": 133, + "tokens_out": 2, + "error": null + }, + { + "id": "route-29", + "set": "route", + "kind": "choice", + "label": "code", + "answer": "code", + "correct": true, + "p_answer": 0.88266, + "p_label": 0.88266, + "distribution": { + "code": 0.88266, + "chat": 0.114203, + "vision": 0.002989, + "long_document": 0.000148 + }, + "wall_ms": 3942, + "server_latency_ms": null, + "tokens_in": 126, + "tokens_out": 2, + "error": null + }, + { + "id": "route-30", + "set": "route", + "kind": "choice", + "label": "chat", + "answer": "chat", + "correct": true, + "p_answer": 0.996279, + "p_label": 0.996279, + "distribution": { + "chat": 0.996279, + "vision": 0.002356, + "code": 0.001152, + "long_document": 0.000212 + }, + "wall_ms": 3936, + "server_latency_ms": null, + "tokens_in": 131, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-01", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.999386, + "p_label": 0.999386, + "distribution": { + "billing": 0.999386, + "technical": 0.000542, + "sales": 6.9e-05, + "account": 3e-06 + }, + "wall_ms": 3932, + "server_latency_ms": null, + "tokens_in": 122, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-02", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.999613, + "p_label": 0.999613, + "distribution": { + "technical": 0.999613, + "billing": 0.000243, + "sales": 0.000133, + "account": 1.1e-05 + }, + "wall_ms": 3926, + "server_latency_ms": null, + "tokens_in": 130, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-03", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.889511, + "p_label": 0.889511, + "distribution": { + "sales": 0.889511, + "billing": 0.106447, + "technical": 0.003843, + "account": 0.000198 + }, + "wall_ms": 3959, + "server_latency_ms": null, + "tokens_in": 123, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-04", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.995859, + "p_label": 0.995859, + "distribution": { + "account": 0.995859, + "technical": 0.003386, + "sales": 0.000612, + "billing": 0.000144 + }, + "wall_ms": 3956, + "server_latency_ms": null, + "tokens_in": 123, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-05", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.993841, + "p_label": 0.993841, + "distribution": { + "billing": 0.993841, + "technical": 0.005686, + "sales": 0.000435, + "account": 3.7e-05 + }, + "wall_ms": 3950, + "server_latency_ms": null, + "tokens_in": 123, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-06", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.999132, + "p_label": 0.999132, + "distribution": { + "technical": 0.999132, + "sales": 0.000462, + "billing": 0.000339, + "account": 6.7e-05 + }, + "wall_ms": 3946, + "server_latency_ms": null, + "tokens_in": 126, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-07", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.990613, + "p_label": 0.990613, + "distribution": { + "sales": 0.990613, + "billing": 0.007409, + "technical": 0.001846, + "account": 0.000132 + }, + "wall_ms": 3921, + "server_latency_ms": null, + "tokens_in": 124, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-08", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.993736, + "p_label": 0.993736, + "distribution": { + "account": 0.993736, + "technical": 0.00276, + "sales": 0.002532, + "billing": 0.000973 + }, + "wall_ms": 3916, + "server_latency_ms": null, + "tokens_in": 122, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-09", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.975135, + "p_label": 0.975135, + "distribution": { + "billing": 0.975135, + "technical": 0.02365, + "sales": 0.000992, + "account": 0.000224 + }, + "wall_ms": 3911, + "server_latency_ms": null, + "tokens_in": 122, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-10", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.995294, + "p_label": 0.995294, + "distribution": { + "technical": 0.995294, + "billing": 0.002131, + "sales": 0.002101, + "account": 0.000474 + }, + "wall_ms": 3906, + "server_latency_ms": null, + "tokens_in": 125, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-11", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.811632, + "p_label": 0.811632, + "distribution": { + "sales": 0.811632, + "billing": 0.183119, + "technical": 0.005025, + "account": 0.000224 + }, + "wall_ms": 3964, + "server_latency_ms": null, + "tokens_in": 121, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-12", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.99902, + "p_label": 0.99902, + "distribution": { + "account": 0.99902, + "technical": 0.000565, + "sales": 0.000364, + "billing": 5.2e-05 + }, + "wall_ms": 3946, + "server_latency_ms": null, + "tokens_in": 128, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-13", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.996794, + "p_label": 0.996794, + "distribution": { + "billing": 0.996794, + "technical": 0.002161, + "sales": 0.000995, + "account": 5e-05 + }, + "wall_ms": 3940, + "server_latency_ms": null, + "tokens_in": 120, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-14", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.999555, + "p_label": 0.999555, + "distribution": { + "technical": 0.999555, + "sales": 0.000236, + "billing": 0.000129, + "account": 8e-05 + }, + "wall_ms": 3936, + "server_latency_ms": null, + "tokens_in": 121, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-15", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.743134, + "p_label": 0.743134, + "distribution": { + "sales": 0.743134, + "technical": 0.249312, + "billing": 0.004511, + "account": 0.003043 + }, + "wall_ms": 4023, + "server_latency_ms": null, + "tokens_in": 122, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-16", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.980534, + "p_label": 0.980534, + "distribution": { + "account": 0.980534, + "technical": 0.01789, + "sales": 0.001137, + "billing": 0.000438 + }, + "wall_ms": 4019, + "server_latency_ms": null, + "tokens_in": 129, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-17", + "set": "triage", + "kind": "choice", + "label": "billing", + "answer": "billing", + "correct": true, + "p_answer": 0.98767, + "p_label": 0.98767, + "distribution": { + "billing": 0.98767, + "sales": 0.009376, + "technical": 0.002904, + "account": 5e-05 + }, + "wall_ms": 4013, + "server_latency_ms": null, + "tokens_in": 124, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-18", + "set": "triage", + "kind": "choice", + "label": "technical", + "answer": "technical", + "correct": true, + "p_answer": 0.999781, + "p_label": 0.999781, + "distribution": { + "technical": 0.999781, + "sales": 0.000154, + "billing": 5.6e-05, + "account": 9e-06 + }, + "wall_ms": 4009, + "server_latency_ms": null, + "tokens_in": 128, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-19", + "set": "triage", + "kind": "choice", + "label": "sales", + "answer": "sales", + "correct": true, + "p_answer": 0.766685, + "p_label": 0.766685, + "distribution": { + "sales": 0.766685, + "billing": 0.226125, + "technical": 0.006954, + "account": 0.000235 + }, + "wall_ms": 4084, + "server_latency_ms": null, + "tokens_in": 121, + "tokens_out": 2, + "error": null + }, + { + "id": "triage-20", + "set": "triage", + "kind": "choice", + "label": "account", + "answer": "account", + "correct": true, + "p_answer": 0.99865, + "p_label": 0.99865, + "distribution": { + "account": 0.99865, + "sales": 0.000828, + "technical": 0.000263, + "billing": 0.000258 + }, + "wall_ms": 4080, + "server_latency_ms": null, + "tokens_in": 121, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-01", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": false, + "correct": false, + "p_answer": 0.986676, + "p_label": 0.013324, + "distribution": { + "false": 0.986676, + "true": 0.013324 + }, + "wall_ms": 4074, + "server_latency_ms": null, + "tokens_in": 103, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-02", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.984281, + "p_label": 0.984281, + "distribution": { + "true": 0.984281, + "false": 0.015719 + }, + "wall_ms": 4055, + "server_latency_ms": null, + "tokens_in": 111, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-03", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.992534, + "p_label": 0.992534, + "distribution": { + "false": 0.992534, + "true": 0.007466 + }, + "wall_ms": 4160, + "server_latency_ms": null, + "tokens_in": 104, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-04", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": false, + "correct": false, + "p_answer": 0.932398, + "p_label": 0.067602, + "distribution": { + "false": 0.932398, + "true": 0.067602 + }, + "wall_ms": 4161, + "server_latency_ms": null, + "tokens_in": 104, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-05", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.98873, + "p_label": 0.98873, + "distribution": { + "false": 0.98873, + "true": 0.01127 + }, + "wall_ms": 4153, + "server_latency_ms": null, + "tokens_in": 104, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-06", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.999225, + "p_label": 0.999225, + "distribution": { + "false": 0.999225, + "true": 0.000775 + }, + "wall_ms": 4149, + "server_latency_ms": null, + "tokens_in": 107, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-07", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.996393, + "p_label": 0.996393, + "distribution": { + "false": 0.996393, + "true": 0.003607 + }, + "wall_ms": 4167, + "server_latency_ms": null, + "tokens_in": 105, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-08", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.995388, + "p_label": 0.995388, + "distribution": { + "false": 0.995388, + "true": 0.004612 + }, + "wall_ms": 4167, + "server_latency_ms": null, + "tokens_in": 103, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-09", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.712851, + "p_label": 0.712851, + "distribution": { + "true": 0.712851, + "false": 0.287149 + }, + "wall_ms": 4160, + "server_latency_ms": null, + "tokens_in": 103, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-10", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.978529, + "p_label": 0.978529, + "distribution": { + "false": 0.978529, + "true": 0.021471 + }, + "wall_ms": 4155, + "server_latency_ms": null, + "tokens_in": 106, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-11", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.979854, + "p_label": 0.979854, + "distribution": { + "false": 0.979854, + "true": 0.020146 + }, + "wall_ms": 4142, + "server_latency_ms": null, + "tokens_in": 102, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-12", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": false, + "correct": false, + "p_answer": 0.553111, + "p_label": 0.446889, + "distribution": { + "false": 0.553111, + "true": 0.446889 + }, + "wall_ms": 4138, + "server_latency_ms": null, + "tokens_in": 109, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-13", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.991068, + "p_label": 0.991068, + "distribution": { + "false": 0.991068, + "true": 0.008932 + }, + "wall_ms": 4133, + "server_latency_ms": null, + "tokens_in": 101, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-14", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": false, + "correct": false, + "p_answer": 0.632473, + "p_label": 0.367527, + "distribution": { + "false": 0.632473, + "true": 0.367527 + }, + "wall_ms": 4128, + "server_latency_ms": null, + "tokens_in": 102, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-15", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.993014, + "p_label": 0.993014, + "distribution": { + "false": 0.993014, + "true": 0.006986 + }, + "wall_ms": 4153, + "server_latency_ms": null, + "tokens_in": 103, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-16", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.908646, + "p_label": 0.908646, + "distribution": { + "true": 0.908646, + "false": 0.091354 + }, + "wall_ms": 4148, + "server_latency_ms": null, + "tokens_in": 110, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-17", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.995814, + "p_label": 0.995814, + "distribution": { + "false": 0.995814, + "true": 0.004186 + }, + "wall_ms": 4143, + "server_latency_ms": null, + "tokens_in": 105, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-18", + "set": "urgency", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.880254, + "p_label": 0.880254, + "distribution": { + "true": 0.880254, + "false": 0.119746 + }, + "wall_ms": 4139, + "server_latency_ms": null, + "tokens_in": 109, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-19", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.986635, + "p_label": 0.986635, + "distribution": { + "false": 0.986635, + "true": 0.013365 + }, + "wall_ms": 4103, + "server_latency_ms": null, + "tokens_in": 102, + "tokens_out": 2, + "error": null + }, + { + "id": "urgency-20", + "set": "urgency", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.990326, + "p_label": 0.990326, + "distribution": { + "false": 0.990326, + "true": 0.009674 + }, + "wall_ms": 4098, + "server_latency_ms": null, + "tokens_in": 102, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-01", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.679296, + "p_label": 0.679296, + "distribution": { + "true": 0.679296, + "false": 0.320704 + }, + "wall_ms": 4093, + "server_latency_ms": null, + "tokens_in": 115, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-02", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.621575, + "p_label": 0.621575, + "distribution": { + "true": 0.621575, + "false": 0.378425 + }, + "wall_ms": 4087, + "server_latency_ms": null, + "tokens_in": 107, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-03", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.898949, + "p_label": 0.898949, + "distribution": { + "false": 0.898949, + "true": 0.101051 + }, + "wall_ms": 4070, + "server_latency_ms": null, + "tokens_in": 107, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-04", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.998839, + "p_label": 0.998839, + "distribution": { + "false": 0.998839, + "true": 0.001161 + }, + "wall_ms": 4065, + "server_latency_ms": null, + "tokens_in": 108, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-05", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.633733, + "p_label": 0.633733, + "distribution": { + "true": 0.633733, + "false": 0.366267 + }, + "wall_ms": 4060, + "server_latency_ms": null, + "tokens_in": 119, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-06", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.990175, + "p_label": 0.990175, + "distribution": { + "false": 0.990175, + "true": 0.009825 + }, + "wall_ms": 4055, + "server_latency_ms": null, + "tokens_in": 113, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-07", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.934643, + "p_label": 0.934643, + "distribution": { + "true": 0.934643, + "false": 0.065357 + }, + "wall_ms": 4052, + "server_latency_ms": null, + "tokens_in": 109, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-08", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.995699, + "p_label": 0.995699, + "distribution": { + "false": 0.995699, + "true": 0.004301 + }, + "wall_ms": 4046, + "server_latency_ms": null, + "tokens_in": 110, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-09", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.745834, + "p_label": 0.745834, + "distribution": { + "true": 0.745834, + "false": 0.254166 + }, + "wall_ms": 4042, + "server_latency_ms": null, + "tokens_in": 108, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-10", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.970229, + "p_label": 0.970229, + "distribution": { + "false": 0.970229, + "true": 0.029771 + }, + "wall_ms": 4036, + "server_latency_ms": null, + "tokens_in": 114, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-11", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.566162, + "p_label": 0.566162, + "distribution": { + "true": 0.566162, + "false": 0.433838 + }, + "wall_ms": 3988, + "server_latency_ms": null, + "tokens_in": 114, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-12", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.99002, + "p_label": 0.99002, + "distribution": { + "false": 0.99002, + "true": 0.00998 + }, + "wall_ms": 3982, + "server_latency_ms": null, + "tokens_in": 116, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-13", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.835931, + "p_label": 0.835931, + "distribution": { + "true": 0.835931, + "false": 0.164069 + }, + "wall_ms": 3978, + "server_latency_ms": null, + "tokens_in": 112, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-14", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.993286, + "p_label": 0.993286, + "distribution": { + "false": 0.993286, + "true": 0.006714 + }, + "wall_ms": 3972, + "server_latency_ms": null, + "tokens_in": 113, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-15", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": false, + "correct": false, + "p_answer": 0.760551, + "p_label": 0.239449, + "distribution": { + "false": 0.760551, + "true": 0.239449 + }, + "wall_ms": 4013, + "server_latency_ms": null, + "tokens_in": 111, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-16", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.824045, + "p_label": 0.824045, + "distribution": { + "false": 0.824045, + "true": 0.175955 + }, + "wall_ms": 4007, + "server_latency_ms": null, + "tokens_in": 120, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-17", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": false, + "correct": false, + "p_answer": 0.543174, + "p_label": 0.456826, + "distribution": { + "false": 0.543174, + "true": 0.456826 + }, + "wall_ms": 4003, + "server_latency_ms": null, + "tokens_in": 109, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-18", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.925629, + "p_label": 0.925629, + "distribution": { + "false": 0.925629, + "true": 0.074371 + }, + "wall_ms": 3997, + "server_latency_ms": null, + "tokens_in": 109, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-19", + "set": "pr_safe", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.62846, + "p_label": 0.62846, + "distribution": { + "true": 0.62846, + "false": 0.37154 + }, + "wall_ms": 4051, + "server_latency_ms": null, + "tokens_in": 107, + "tokens_out": 2, + "error": null + }, + { + "id": "pr_safe-20", + "set": "pr_safe", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.99483, + "p_label": 0.99483, + "distribution": { + "false": 0.99483, + "true": 0.00517 + }, + "wall_ms": 4045, + "server_latency_ms": null, + "tokens_in": 114, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-01", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.95361, + "p_label": 0.95361, + "distribution": { + "true": 0.95361, + "false": 0.04639 + }, + "wall_ms": 4041, + "server_latency_ms": null, + "tokens_in": 74, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-02", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.999397, + "p_label": 0.999397, + "distribution": { + "false": 0.999397, + "true": 0.000603 + }, + "wall_ms": 4035, + "server_latency_ms": null, + "tokens_in": 75, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-03", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.994192, + "p_label": 0.994192, + "distribution": { + "true": 0.994192, + "false": 0.005808 + }, + "wall_ms": 4089, + "server_latency_ms": null, + "tokens_in": 65, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-04", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.990694, + "p_label": 0.990694, + "distribution": { + "true": 0.990694, + "false": 0.009306 + }, + "wall_ms": 4083, + "server_latency_ms": null, + "tokens_in": 67, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-05", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.618266, + "p_label": 0.618266, + "distribution": { + "true": 0.618266, + "false": 0.381734 + }, + "wall_ms": 4078, + "server_latency_ms": null, + "tokens_in": 80, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-06", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.998904, + "p_label": 0.998904, + "distribution": { + "false": 0.998904, + "true": 0.001096 + }, + "wall_ms": 4073, + "server_latency_ms": null, + "tokens_in": 74, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-07", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.997242, + "p_label": 0.997242, + "distribution": { + "false": 0.997242, + "true": 0.002758 + }, + "wall_ms": 4029, + "server_latency_ms": null, + "tokens_in": 71, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-08", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.997796, + "p_label": 0.997796, + "distribution": { + "true": 0.997796, + "false": 0.002204 + }, + "wall_ms": 4024, + "server_latency_ms": null, + "tokens_in": 74, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-09", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.995258, + "p_label": 0.995258, + "distribution": { + "true": 0.995258, + "false": 0.004742 + }, + "wall_ms": 4019, + "server_latency_ms": null, + "tokens_in": 67, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-10", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.900869, + "p_label": 0.900869, + "distribution": { + "false": 0.900869, + "true": 0.099131 + }, + "wall_ms": 4014, + "server_latency_ms": null, + "tokens_in": 71, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-11", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.940458, + "p_label": 0.940458, + "distribution": { + "true": 0.940458, + "false": 0.059542 + }, + "wall_ms": 3948, + "server_latency_ms": null, + "tokens_in": 79, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-12", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.979713, + "p_label": 0.979713, + "distribution": { + "true": 0.979713, + "false": 0.020287 + }, + "wall_ms": 3943, + "server_latency_ms": null, + "tokens_in": 71, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-13", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.997558, + "p_label": 0.997558, + "distribution": { + "false": 0.997558, + "true": 0.002442 + }, + "wall_ms": 3938, + "server_latency_ms": null, + "tokens_in": 68, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-14", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.644133, + "p_label": 0.644133, + "distribution": { + "true": 0.644133, + "false": 0.355867 + }, + "wall_ms": 3932, + "server_latency_ms": null, + "tokens_in": 70, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-15", + "set": "fact", + "kind": "noul", + "label": false, + "answer": true, + "correct": false, + "p_answer": 0.63427, + "p_label": 0.36573, + "distribution": { + "true": 0.63427, + "false": 0.36573 + }, + "wall_ms": 3993, + "server_latency_ms": null, + "tokens_in": 68, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-16", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.584615, + "p_label": 0.584615, + "distribution": { + "true": 0.584615, + "false": 0.415385 + }, + "wall_ms": 3988, + "server_latency_ms": null, + "tokens_in": 77, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-17", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.999279, + "p_label": 0.999279, + "distribution": { + "false": 0.999279, + "true": 0.000721 + }, + "wall_ms": 3990, + "server_latency_ms": null, + "tokens_in": 69, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-18", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.656469, + "p_label": 0.656469, + "distribution": { + "true": 0.656469, + "false": 0.343531 + }, + "wall_ms": 3981, + "server_latency_ms": null, + "tokens_in": 78, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-19", + "set": "fact", + "kind": "noul", + "label": false, + "answer": false, + "correct": true, + "p_answer": 0.998465, + "p_label": 0.998465, + "distribution": { + "false": 0.998465, + "true": 0.001535 + }, + "wall_ms": 3549, + "server_latency_ms": null, + "tokens_in": 66, + "tokens_out": 2, + "error": null + }, + { + "id": "fact-20", + "set": "fact", + "kind": "noul", + "label": true, + "answer": true, + "correct": true, + "p_answer": 0.985798, + "p_label": 0.985798, + "distribution": { + "true": 0.985798, + "false": 0.014202 + }, + "wall_ms": 3544, + "server_latency_ms": null, + "tokens_in": 68, + "tokens_out": 2, + "error": null + } + ] + }, + "notes": [ + "Measured by scripts/ainode-bench.py decide in 88s over 110 labeled items from items.json (decide-110); nothing was loaded, unloaded or restarted.", + "Accuracy is the weakest number here: the failure mode automation cares about is a wrong answer at high confidence, which is what the Brier score, the calibration error and the wrong-at-threshold counts measure.", + "Confidence is the probability the backend put on the answer it gave. Brier is a single term on the labeled option's probability, and the calibration error uses 5 bins over the reliability table in the same block.", + "Cost is $0 for the chat backend: it runs on our own hardware and nobody bills per token for it. The electricity is real and is not a number this record claims to have measured.", + "The chat backend is the fallback instrument: lettered options with thinking off, and a distribution softmaxed from the top logprobs of the one letter token. Those probabilities are over letters rather than over meanings, so they are a weaker statement than a decision endpoint's." + ], + "source": "scripts/ainode-bench.py decide" +} diff --git a/bench/results/20261002-010753-nemotron-nano-4b-q4km-v100-pollux-one-v100-32gb-llama_cpp-server-cuda-q4_k_m-outside-ainode.json b/bench/results/20261002-010753-nemotron-nano-4b-q4km-v100-pollux-one-v100-32gb-llama_cpp-server-cuda-q4_k_m-outside-ainode.json new file mode 100644 index 00000000..54d9e697 --- /dev/null +++ b/bench/results/20261002-010753-nemotron-nano-4b-q4km-v100-pollux-one-v100-32gb-llama_cpp-server-cuda-q4_k_m-outside-ainode.json @@ -0,0 +1,79 @@ +{ + "schema": 1, + "stamp": "20261002-010753", + "label": "pollux one v100 32gb, llama.cpp server-cuda q4_k_m, outside ainode", + "model": { + "id": "nemotron-nano-4b-q4km-v100" + }, + "placement": { + "engine": "vllm" + }, + "settings": { + "thinking": false, + "max_tokens": 200, + "temperature": null, + "sustained_tokens": 1500, + "reasoning_tokens": 600, + "depths_requested": [ + 4000, + 16000, + 32000, + 64000, + 120000 + ], + "streams_requested": [ + 1, + 2, + 4 + ], + "chars_per_token_probe": 5.38 + }, + "results": { + "single_stream": { + "decode_tok_s": 152.58, + "gen_tokens": 200, + "ttft_ms": 657, + "prompt_tokens": 78 + }, + "concurrency": [ + { + "streams": 1, + "aggregate_tok_s": 106.7, + "ok": 1, + "failed": 0, + "wall_s": 1.67, + "per_stream_tok_s": 153.2, + "median_ttft_ms": 513 + }, + { + "streams": 2, + "aggregate_tok_s": 164.5, + "ok": 2, + "failed": 0, + "wall_s": 2.4, + "per_stream_tok_s": 119.4, + "median_ttft_ms": 736 + }, + { + "streams": 4, + "aggregate_tok_s": 192.0, + "ok": 4, + "failed": 0, + "wall_s": 3.95, + "per_stream_tok_s": 79.4, + "median_ttft_ms": 1431 + } + ], + "sustained": { + "decode_tok_s": 152.65, + "gen_tokens": 831, + "wall_s": 6.722, + "ttft_ms": 1284 + } + }, + "notes": [ + "Measured by scripts/ainode-bench.py in 17s; nothing loaded, unloaded or restarted.", + "Sections other than reasoning_tax ran with enable_thinking=false." + ], + "source": "scripts/ainode-bench.py" +} diff --git a/bench/results/20261002-010806-nemotron-nano-4b-v100-pollux-one-v100-32gb-onecat-vllm-src-full-fp16-outside-ainode.json b/bench/results/20261002-010806-nemotron-nano-4b-v100-pollux-one-v100-32gb-onecat-vllm-src-full-fp16-outside-ainode.json new file mode 100644 index 00000000..6dcfb761 --- /dev/null +++ b/bench/results/20261002-010806-nemotron-nano-4b-v100-pollux-one-v100-32gb-onecat-vllm-src-full-fp16-outside-ainode.json @@ -0,0 +1,90 @@ +{ + "schema": 1, + "stamp": "20261002-010806", + "label": "pollux one v100 32gb, onecat-vllm src-full fp16, outside ainode", + "model": { + "id": "nemotron-nano-4b-v100" + }, + "placement": { + "engine": "vllm", + "max_model_len": 32768 + }, + "settings": { + "thinking": false, + "max_tokens": 200, + "temperature": null, + "sustained_tokens": 1500, + "reasoning_tokens": 600, + "depths_requested": [ + 4000, + 16000, + 32000, + 64000, + 120000 + ], + "streams_requested": [ + 1, + 2, + 4, + 8 + ], + "chars_per_token_probe": 5.38 + }, + "results": { + "single_stream": { + "decode_tok_s": 83.8, + "gen_tokens": 189, + "ttft_ms": 121, + "prompt_tokens": 83 + }, + "concurrency": [ + { + "streams": 1, + "aggregate_tok_s": 80.1, + "ok": 1, + "failed": 0, + "wall_s": 2.28, + "per_stream_tok_s": 83.8, + "median_ttft_ms": 113 + }, + { + "streams": 2, + "aggregate_tok_s": 43.6, + "ok": 2, + "failed": 0, + "wall_s": 9.17, + "per_stream_tok_s": 74.5, + "median_ttft_ms": 6490 + }, + { + "streams": 4, + "aggregate_tok_s": 244.0, + "ok": 4, + "failed": 0, + "wall_s": 2.95, + "per_stream_tok_s": 72.2, + "median_ttft_ms": 217 + }, + { + "streams": 8, + "aggregate_tok_s": 463.2, + "ok": 8, + "failed": 0, + "wall_s": 3.27, + "per_stream_tok_s": 66.8, + "median_ttft_ms": 311 + } + ], + "sustained": { + "decode_tok_s": 83.33, + "gen_tokens": 1153, + "wall_s": 13.939, + "ttft_ms": 114 + } + }, + "notes": [ + "Measured by scripts/ainode-bench.py in 34s; nothing loaded, unloaded or restarted.", + "Sections other than reasoning_tax ran with enable_thinking=false." + ], + "source": "scripts/ainode-bench.py" +} diff --git a/bench/results/20261002-010827-nemotron-nano-4b-v100-pollux-one-v100-32gb-onecat-vllm-src-full-fp16-outside-ainode-agentic.json b/bench/results/20261002-010827-nemotron-nano-4b-v100-pollux-one-v100-32gb-onecat-vllm-src-full-fp16-outside-ainode-agentic.json new file mode 100644 index 00000000..e01f34f1 --- /dev/null +++ b/bench/results/20261002-010827-nemotron-nano-4b-v100-pollux-one-v100-32gb-onecat-vllm-src-full-fp16-outside-ainode-agentic.json @@ -0,0 +1,298 @@ +{ + "schema": 1, + "stamp": "20261002-010827", + "label": "pollux one v100 32gb, onecat-vllm src-full fp16, outside ainode", + "model": { + "id": "nemotron-nano-4b-v100" + }, + "placement": {}, + "settings": { + "groups": [ + "A", + "B", + "C", + "D", + "E", + "F", + "G" + ], + "needle_tokens": [ + 8000 + ], + "quick": false, + "temperature": 1.0, + "timeout_s": 900, + "thinking_switch": "enable_thinking", + "probes_requested": 22 + }, + "agentic": { + "score": { + "pass": 1, + "total": 22 + }, + "groups": { + "A": { + "pass": 0, + "total": 4 + }, + "B": { + "pass": 0, + "total": 4 + }, + "C": { + "pass": 0, + "total": 3 + }, + "D": { + "pass": 0, + "total": 4 + }, + "E": { + "pass": 0, + "total": 1 + }, + "F": { + "pass": 1, + "total": 1 + }, + "G": { + "pass": 0, + "total": 5 + } + }, + "endpoint": "http://100.124.227.93:8100/v1", + "protocol": { + "groups": [ + "A", + "B", + "C", + "D", + "E", + "F", + "G" + ], + "needle_tokens": [ + 8000 + ], + "temperature": 1.0, + "timeout_s": 900, + "thinking_switch": "enable_thinking", + "max_turns": 4, + "verdicts": "mechanical; group C is executed, group G is a judged tool trace" + }, + "probes": [ + { + "id": "A1_format", + "group": "A", + "pass": false, + "wall_s": 0.51, + "completion_tokens": 35, + "note": "0 line(s), wanted 5", + "excerpt": "" + }, + { + "id": "A2_json_only", + "group": "A", + "pass": false, + "wall_s": 0.36, + "completion_tokens": 22, + "note": "not raw JSON: Expecting value: line 1 column 1 (char 0)", + "excerpt": "" + }, + { + "id": "A3_constraints", + "group": "A", + "pass": false, + "wall_s": 0.61, + "completion_tokens": 43, + "note": "0 full stop(s), wanted about 3 sentences", + "excerpt": "" + }, + { + "id": "A4_persona", + "group": "A", + "pass": false, + "wall_s": 0.94, + "completion_tokens": 71, + "note": "dropped the persona: no trailing 'Arr!'", + "excerpt": "" + }, + { + "id": "B1_tool_single", + "group": "B", + "pass": false, + "wall_s": 0.54, + "completion_tokens": 35, + "note": "0 call(s), wanted 1: no calls", + "excerpt": "" + }, + { + "id": "B2_tool_parallel", + "group": "B", + "pass": false, + "wall_s": 1.51, + "completion_tokens": 115, + "note": "0 call(s), wanted 3: no calls", + "excerpt": "" + }, + { + "id": "B3_tool_not_needed", + "group": "B", + "pass": false, + "wall_s": 0.4, + "completion_tokens": 23, + "note": "no tool call and no text either", + "excerpt": "" + }, + { + "id": "B4_tool_roundtrip", + "group": "B", + "pass": false, + "wall_s": 0.54, + "completion_tokens": 35, + "note": "no tool call to answer from", + "excerpt": "" + }, + { + "id": "C1_ttl_cache", + "group": "C", + "pass": false, + "wall_s": 3.92, + "completion_tokens": 319, + "note": "NameError: name 'TTLCache' is not defined", + "excerpt": "" + }, + { + "id": "C2_intervals", + "group": "C", + "pass": false, + "wall_s": 3.68, + "completion_tokens": 299, + "note": "NameError: name 'merge_intervals' is not defined", + "excerpt": "" + }, + { + "id": "C3_bugfix", + "group": "C", + "pass": false, + "wall_s": 1.03, + "completion_tokens": 78, + "note": "NameError: name 'top_k_words' is not defined", + "excerpt": "" + }, + { + "id": "D1_decimal", + "group": "D", + "pass": false, + "wall_s": 0.15, + "completion_tokens": 4, + "note": "wanted 9.9", + "excerpt": "" + }, + { + "id": "D2_strawberry", + "group": "D", + "pass": false, + "wall_s": 0.12, + "completion_tokens": 2, + "note": "wanted 3", + "excerpt": "" + }, + { + "id": "D3_sisters", + "group": "D", + "pass": false, + "wall_s": 0.12, + "completion_tokens": 2, + "note": "wanted 2", + "excerpt": "" + }, + { + "id": "D4_knights", + "group": "D", + "pass": false, + "wall_s": 0.22, + "completion_tokens": 10, + "note": "wanted A=Knight, B=Knave", + "excerpt": "" + }, + { + "id": "E_needle_8000", + "group": "E", + "pass": false, + "wall_s": 1.42, + "completion_tokens": 11, + "note": "did not find the password; prompt_tokens=6516", + "excerpt": "" + }, + { + "id": "F1_thinking_off", + "group": "F", + "pass": true, + "wall_s": 0.15, + "completion_tokens": 4, + "note": "answered with no reasoning channel", + "excerpt": "391" + }, + { + "id": "G1_tool_chain", + "group": "G", + "pass": false, + "wall_s": 0.28, + "completion_tokens": 14, + "note": "answered without calling a tool", + "excerpt": "" + }, + { + "id": "G2_tool_error", + "group": "G", + "pass": false, + "wall_s": 0.58, + "completion_tokens": 38, + "note": "never called the weather tool", + "excerpt": "" + }, + { + "id": "G3_arg_schema", + "group": "G", + "pass": false, + "wall_s": 0.68, + "completion_tokens": 47, + "note": "0 call(s), wanted 1: no calls", + "excerpt": "[]" + }, + { + "id": "G4_structured", + "group": "G", + "pass": false, + "wall_s": 2.2, + "completion_tokens": 45, + "note": "no JSON object in the reply [json_schema]", + "excerpt": "" + }, + { + "id": "G5_persistence", + "group": "G", + "pass": false, + "wall_s": 0.89, + "completion_tokens": 41, + "note": "0/4 turns ended with DONE; missed turn(s) [1, 2, 3, 4]", + "excerpt": "" + } + ], + "needle": { + "8000": false + }, + "thinking_off_supported": true, + "vision_supported": null, + "structured_output_mode": "json_schema" + }, + "notes": [ + "Measured by scripts/ainode-bench.py agentic in 21s; nothing loaded, unloaded or restarted.", + "Every verdict is mechanical: no judge model and no reply was read by a person to reach a score.", + "Group C runs the model's own code against asserts it never saw, in a subprocess on the machine driving the bench.", + "Group G drives real tool loops and judges the call trace: the order of the calls, the arguments, and whether a stated fact came back from a tool.", + "Groups run: A, B, C, D, E, F, G. A group that did not run is absent from the score rather than counted as zero." + ], + "source": "scripts/ainode-bench.py agentic" +} diff --git a/bench/results/20261002-010855-nemotron-nano-4b-q4km-v100-pollux-one-v100-32gb-llama_cpp-server-cuda-q4_k_m-outside-ainode-agentic.json b/bench/results/20261002-010855-nemotron-nano-4b-q4km-v100-pollux-one-v100-32gb-llama_cpp-server-cuda-q4_k_m-outside-ainode-agentic.json new file mode 100644 index 00000000..3a6921c9 --- /dev/null +++ b/bench/results/20261002-010855-nemotron-nano-4b-q4km-v100-pollux-one-v100-32gb-llama_cpp-server-cuda-q4_k_m-outside-ainode-agentic.json @@ -0,0 +1,298 @@ +{ + "schema": 1, + "stamp": "20261002-010855", + "label": "pollux one v100 32gb, llama.cpp server-cuda q4_k_m, outside ainode", + "model": { + "id": "nemotron-nano-4b-q4km-v100" + }, + "placement": {}, + "settings": { + "groups": [ + "A", + "B", + "C", + "D", + "E", + "F", + "G" + ], + "needle_tokens": [ + 8000 + ], + "quick": false, + "temperature": 1.0, + "timeout_s": 900, + "thinking_switch": "enable_thinking", + "probes_requested": 22 + }, + "agentic": { + "score": { + "pass": 20, + "total": 22 + }, + "groups": { + "A": { + "pass": 3, + "total": 4 + }, + "B": { + "pass": 3, + "total": 4 + }, + "C": { + "pass": 3, + "total": 3 + }, + "D": { + "pass": 4, + "total": 4 + }, + "E": { + "pass": 1, + "total": 1 + }, + "F": { + "pass": 1, + "total": 1 + }, + "G": { + "pass": 5, + "total": 5 + } + }, + "endpoint": "http://100.124.227.93:8101/v1", + "protocol": { + "groups": [ + "A", + "B", + "C", + "D", + "E", + "F", + "G" + ], + "needle_tokens": [ + 8000 + ], + "temperature": 1.0, + "timeout_s": 900, + "thinking_switch": "enable_thinking", + "max_turns": 4, + "verdicts": "mechanical; group C is executed, group G is a judged tool trace" + }, + "probes": [ + { + "id": "A1_format", + "group": "A", + "pass": true, + "wall_s": 5.31, + "completion_tokens": 736, + "note": "5 lines, names in caps, numbering loose", + "excerpt": "N. JUPITER N. SATURN N. URANUS N. NEPTUNE N. EARTH" + }, + { + "id": "A2_json_only", + "group": "A", + "pass": false, + "wall_s": 1.31, + "completion_tokens": 122, + "note": "landlocked is False, Austria is landlocked", + "excerpt": "{\"city\":\"Vienna\",\"population_millions\":1.9,\"landlocked\":false}" + }, + { + "id": "A3_constraints", + "group": "A", + "pass": true, + "wall_s": 3.99, + "completion_tokens": 531, + "note": "no z, no 'the', 3 full stops", + "excerpt": "Handlebars hold rider. Pedal motion turns wheel axle. Air pressure pushes tire forward." + }, + { + "id": "A4_persona", + "group": "A", + "pass": true, + "wall_s": 3.86, + "completion_tokens": 515, + "note": "stayed in character", + "excerpt": "Ahoy, matey! A mutex be a special tool that makes sure one thread use the shared resource at one time, like a chest, so no treasure be stolen by the wrong pirate. Arr!" + }, + { + "id": "B1_tool_single", + "group": "B", + "pass": true, + "wall_s": 1.03, + "completion_tokens": 73, + "note": "get_weather({\"city\":\"Tokyo\",\"unit\":\"c\"})", + "excerpt": "" + }, + { + "id": "B2_tool_parallel", + "group": "B", + "pass": false, + "wall_s": 0.96, + "completion_tokens": 106, + "note": "1 call(s), wanted 3: get_weather({\"city\":\"Paris\",\"unit\":\"c\"})", + "excerpt": "" + }, + { + "id": "B3_tool_not_needed", + "group": "B", + "pass": true, + "wall_s": 4.42, + "completion_tokens": 632, + "note": "answered without a tool", + "excerpt": "Leaves whisper low now Crisp air kisses the path home Autumn's breath sighs low" + }, + { + "id": "B4_tool_roundtrip", + "group": "B", + "pass": true, + "wall_s": 2.0, + "completion_tokens": 183, + "note": "used the tool result", + "excerpt": "The current weather in Tokyo is **31\u00b0C** with **thunderstorms**." + }, + { + "id": "C1_ttl_cache", + "group": "C", + "pass": true, + "wall_s": 8.48, + "completion_tokens": 1209, + "note": "hidden asserts passed", + "excerpt": "```python import time from collections import OrderedDict class TTLCache: def __init__(self, capacity: int, ttl: int, clock=lambda: time.time): self.cap = capacity self.ttl = ttl self.clock = clock self.data = OrderedDict() # key -> (value, expire_time) def get(self, key): now = self.clock() if key ..." + }, + { + "id": "C2_intervals", + "group": "C", + "pass": true, + "wall_s": 4.94, + "completion_tokens": 680, + "note": "hidden asserts passed", + "excerpt": "```python def merge_intervals(intervals): \"\"\" Merge overlapping or touching [start, end] intervals. Returns a sorted list of merged intervals. \"\"\" if not intervals: return [] # sort by start time intervals = sorted(intervals, key=lambda x: x[0]) merged = [intervals[0]] for cur_start, cur_end in inte..." + }, + { + "id": "C3_bugfix", + "group": "C", + "pass": true, + "wall_s": 2.15, + "completion_tokens": 249, + "note": "hidden asserts passed", + "excerpt": "```python import re from collections import Counter def top_k_words(text, k): words = re.findall(r'[a-z]+', text.lower()) counts = Counter(words) sorted_items = sorted(counts.items(), key=lambda kv: (-kv[1], kv[0])) return [w for w, c in sorted_items[:k]] ```" + }, + { + "id": "D1_decimal", + "group": "D", + "pass": true, + "wall_s": 1.12, + "completion_tokens": 93, + "note": "said 9.9", + "excerpt": "9.9" + }, + { + "id": "D2_strawberry", + "group": "D", + "pass": true, + "wall_s": 0.95, + "completion_tokens": 67, + "note": "said 3", + "excerpt": "3" + }, + { + "id": "D3_sisters", + "group": "D", + "pass": true, + "wall_s": 1.54, + "completion_tokens": 160, + "note": "said 2", + "excerpt": "2" + }, + { + "id": "D4_knights", + "group": "D", + "pass": true, + "wall_s": 3.79, + "completion_tokens": 504, + "note": "said A=Knight, B=Knave", + "excerpt": "A=Knight, B=Knave." + }, + { + "id": "E_needle_8000", + "group": "E", + "pass": true, + "wall_s": 3.47, + "completion_tokens": 215, + "note": "found the password; prompt_tokens=6517", + "excerpt": "violet-otter-2931" + }, + { + "id": "F1_thinking_off", + "group": "F", + "pass": true, + "wall_s": 0.74, + "completion_tokens": 4, + "note": "answered with no reasoning channel", + "excerpt": "391" + }, + { + "id": "G1_tool_chain", + "group": "G", + "pass": true, + "wall_s": 2.9, + "completion_tokens": 273, + "note": "list_files -> read_file; answered 137", + "excerpt": "137" + }, + { + "id": "G2_tool_error", + "group": "G", + "pass": true, + "wall_s": 2.35, + "completion_tokens": 166, + "note": "called the tool with 'San Francisco' after the error", + "excerpt": "The current temperature in San Francisco is **18 \u00b0C**." + }, + { + "id": "G3_arg_schema", + "group": "G", + "pass": true, + "wall_s": 1.55, + "completion_tokens": 127, + "note": "mode='daily' retries=3 notify=True", + "excerpt": "[[\"schedule_job\", \"{\\\"mode\\\":\\\"daily\\\",\\\"retries\\\":3,\\\"notify\\\":true}\"]]" + }, + { + "id": "G4_structured", + "group": "G", + "pass": true, + "wall_s": 1.24, + "completion_tokens": 110, + "note": "four keys, right types, severity=2 [json_schema]", + "excerpt": "{ \"title\": \"Login Service 500 Errors After Bad Deploy\", \"severity\": 2, \"resolved\": true, \"tags\": [\"login\", \"deploy\"] }" + }, + { + "id": "G5_persistence", + "group": "G", + "pass": true, + "wall_s": 4.36, + "completion_tokens": 435, + "note": "all 4 turns ended with DONE", + "excerpt": "You asked: \"What is the capital of France?\" DONE" + } + ], + "needle": { + "8000": true + }, + "thinking_off_supported": true, + "vision_supported": null, + "structured_output_mode": "json_schema" + }, + "notes": [ + "Measured by scripts/ainode-bench.py agentic in 63s; nothing loaded, unloaded or restarted.", + "Every verdict is mechanical: no judge model and no reply was read by a person to reach a score.", + "Group C runs the model's own code against asserts it never saw, in a subprocess on the machine driving the bench.", + "Group G drives real tool loops and judges the call trace: the order of the calls, the arguments, and whether a stated fact came back from a tool.", + "Groups run: A, B, C, D, E, F, G. A group that did not run is absent from the score rather than counted as zero." + ], + "source": "scripts/ainode-bench.py agentic" +}