Repository navigation
Citation, and bidirectional attention results on a smaller decoder - #5
Merged
Merged
Conversation
…Qwen3-0.6B - README ends with a BibTeX entry; CITATION.cff (valid CFF 1.2.0) gives GitHub's "Cite this repository". - docs/benchmarks.md: 50-step comparison on deciformer-15k with Qwen3-0.6B: causal 0.672, bidirectional with the last-token readout 0.449, with a mean pooler 0.441. The README's attention section quotes it.
akanyaani
added a commit
that referenced
this pull request
Oct 23, 2026
Pre-release fixes from the developer review (includes #5)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
CITATION.cff(validated against CFF 1.2.0) makes GitHub show a "Cite this repository" button. The author is listed as FrontiersMind AI, matching the package metadata; individual names can be added.deciformer15k_qwen3recipe (Qwen3-0.6B, deciformer-15k, 256 dev records), identical except for attention and pooling:The weak bidirectional result is not specific to Qwen3.5: a causally pretrained decoder switched to bidirectional attention starts from a much worse point and learns far more slowly. The results are added to
docs/benchmarks.md, and the README's attention section quotes them.