Use full-population data for output weights#51
Merged
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
MaxGhenis
added a commit
that referenced
this pull request
May 19, 2026
The pr-46-review-era stack landed before PR #51 ("Use full-population data for output weights") and Daphne's PR #47 ACA-from-scoring removal. The result on main right now: - ruff format has drifted on policybench/analysis.py, policybench/full_run_export.py, and tests/test_spec.py - policybench/population_weights.json still contains premium_tax_credit, but the headline output set no longer does (post-#47) That makes both lint and test red on every PR against current main. This commit: - Reformats the three drifted files with ruff format - Regenerates policybench/population_weights.json via policybench population-weights, which drops the now-out-of-scope premium_tax_credit entry - Updates the matching SHA-256 in paper/snapshot/20260501/manifest.json so test_snapshot_manifest_hashes_match_population_weight_artifact passes against the new artifact Verified: uv run pytest -m "not slow" -q → 258 passed; uv run ruff format --check . → clean. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
MaxGhenis
added a commit
that referenced
this pull request
May 19, 2026
* Hero: replace inventory subtitle with a mission line The hero used to read "14 models on 100 households across 18 tax and benefit outputs." That information is already in the stats chip row below (Models / Households / Outputs / snapshot date), so the subtitle was redundant inventory rather than orienting copy. Replace with a one-line mission statement that says what the benchmark is actually for: "Testing how accurately language models calculate household taxes and benefits." The 100% = within-1% explainer continues immediately after as the muted continuation, so readers still learn what the score means. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * Clean PolicyBench deployment config * Fix CI: ruff format, regenerate population-weight artifact The pr-46-review-era stack landed before PR #51 ("Use full-population data for output weights") and Daphne's PR #47 ACA-from-scoring removal. The result on main right now: - ruff format has drifted on policybench/analysis.py, policybench/full_run_export.py, and tests/test_spec.py - policybench/population_weights.json still contains premium_tax_credit, but the headline output set no longer does (post-#47) That makes both lint and test red on every PR against current main. This commit: - Reformats the three drifted files with ruff format - Regenerates policybench/population_weights.json via policybench population-weights, which drops the now-out-of-scope premium_tax_credit entry - Updates the matching SHA-256 in paper/snapshot/20260501/manifest.json so test_snapshot_manifest_hashes_match_population_weight_artifact passes against the new artifact Verified: uv run pytest -m "not slow" -q → 258 passed; uv run ruff format --check . → clean. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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.
Summary
Fixes #49.
Tests