Skip to content

docs: align clip and rubric examples with their commands #19

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
HomenShum merged 1 commit into from
Oct 7, 2026
Merged

docs: align clip and rubric examples with their commands #19

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 8 additions & 8 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -223,8 +223,8 @@ what problem it solved, or how to point it at a question of their own.
**Comprehension is scored from a named audience's seat** — that is what `--for` is:

```bash
npm run clip -- --comp WTC-TShero --out out/trialscope.mp4 --for "a non-technical person who has never heard of this domain"
npm run judge -- out/trialscope.mp4 --for "a frontend engineer evaluating adoption" --gate 28
npm run iterate -- --comp WTC-TShero --out out/trialscope.mp4 --for "a non-technical person who has never heard of this domain"
npm run judge:rubric -- out/trialscope.mp4 --for "a frontend engineer evaluating adoption" --gate 28
```

The same cut is a 2 on `lay_sense` for a domain expert and a 0 for someone who has
Expand All @@ -243,12 +243,12 @@ own critic's brief converges on whatever the critic likes, which is not the same
a good demo, and leaves nobody holding the taste. The brief is written to disk and
the process exits non-zero; a human or an agent applies it; round N+1 begins.

**Anti-uniformity is enforced in code, not asked for in the prompt.** The rubric
carried an anti-uniformity clause for three revisions and the judge still returned
1/2 on 18 of 20 dimensions — a description wearing a score's clothes. Now if one
score covers >70% of dimensions the judgement is re-requested once, with the
offending distribution quoted back. A gate that returns the same verdict for every
input is not a gate, and that includes the flat-1 verdict.
**Anti-uniformity re-asking is opt-in.** `npm run judge:rubric` defaults to three
samples; add `--reask` to request one extra judgment if more than 70% of dimensions
share a single score. Re-asking is off by default because the historical comparison
above reported downward bias. `npm run iterate` calls this numeric rubric;
`npm run clip` calls the separate video judge through its own gates. These command
paths use different rubrics; use the flags documented for the chosen path.

The same three files are vendored into [NodeVideo](https://github.com/HomenShum/NodeVideo)
and [NodeSlide](https://github.com/HomenShum/NodeSlide) under `tools/clip-gate/`,
Expand Down
Loading