bench(coverage): figure-coverage metric + baseline (73% on the sample)#140
Merged
Conversation
A deterministic, extraction-independent yardstick for "did figmark capture the figures the document declares?". Ground truth = the document's own numbered captions (Chart/Figure/Diagram N); a number is covered if a non-skipped figure lands on a page where that caption appears. Page-level, so it is a conservative lower bound on misses — it never cries wolf. Baseline on the sample (current main extraction): BoC 58%, BoJ 79%, aggregate 73%. Misses are not one chart type (BoC Chart 4 is a line chart dropped by the MIN_SOLID_DRAWINGS gate). This anchors the coming extraction redesign (drop the geometric pre-classification). Quality/relevance is a separate track (LLM judge). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Starts the eval effort at the coverage end, per the design discussion: a deterministic, extraction-independent metric for did figmark capture the figures the document declares?
Method
Chart/Figure/Diagram N) + their pages.figures.json+ their pages.Baseline (current
mainextraction)Misses are not one chart type — BoC Chart 4 is a plain line chart dropped by the
MIN_SOLID_DRAWINGS_PER_CLUSTERgate. This is the number to beat when we replace the geometric pre-classification with "render every visual region, let the vision model decide."scripts/coverage_bench/(script + committedBASELINE.md), matching the existing*_bench/convention. Scripts-only.🤖 Generated with Claude Code