Refusal-justify T2: rotate spent holdout -> regression (9) + fresh blind holdout#28
Merged
Conversation
…efusal-justify T2 / gap #3) - Extend REGRESSION_LABELS from 6 -> 9 entries (reg_medical_coercive: True, reg_swim_cooperative: False, reg_religion_hostile: False) - Add test_reg_medical_coercive_now_completes (gap #3 fix verified via T1) - Add test_new_reg_negatives_gap3_stay_non_coercive (specificity guard) - Update test_regression_labels to 9-entry dict - Update tiered confusion assertion to (tp=3, fp=0, fn=0, tn=6) - Wire data/regression/ new files and fresh data/holdout/ files All 262 tests pass. REGRESSION 9/9. reg_religion_hostile 4 hostile hits, 0 FP.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Refusal+justify vocabulary iteration (recall gap #3), Task 2 of 2.
Rotates the spent blind holdout (which drove the gap-#3 fix) into the asserted regression set.
data/holdout/hold_*.json→data/regression/reg_{medical_coercive,swim_cooperative,religion_hostile}.json(byte-identical blobs).REGRESSION_LABELS→ 9 entries. Asserts:reg_medical_coercivenow completes (the gap-ci: add mypy type gate (pragmatic) + type fixes (Phase 0 A2) #3 target, coercer Derek, cycles=2); the 2 new negatives + the existing 6 stay correct; regression confusion(tp=3, fp=0, fn=0, tn=6).demo --evalprints all 4 tiers (9 regression corpora).Fresh-holdout generalization finding (surfaced, not chased): the new
hold_coercive(a relocation dispute) is itself a false negative — 3 stage-hits, 0 envelopes (= gap #4). Specificity held at 1.0 (both negatives correct). Per discipline the corpus was NOT edited — signal for a future iteration.🤖 Generated with Claude Code