Unit 6 (prompt-design): split criterion 2 (MEDIUM, preserve-by-default)#164
Merged
Conversation
…-default) Per docs/RUBRIC_AUDIT.md (MEDIUM): old c2 bundled 'names a concrete failure mode' with 'explains the mechanism.' Splits into name-the-failure-mode c2 and a new c3 (explain the mechanism); renumbers regime distinction to position 4. Rubric grows 3 -> 4. Preserve-by-default (docs/REGRESSION_GATE.md): faithful decomposition of the locked Opus values — old-c2=T → c2=T,c3=T. p007 is the lone c2=T/c3=F differential (its authored label says it names 'instructions miss edge cases' without the ambiguous-criteria mechanism); p009/p011 name no failure mode → c2=F,c3=F. No realignments, no judgment-flips. c4 (regime) carries old-c3 unchanged. Updated p007/p011 labels for the 4-criterion shape. Sonnet gate disagreements are documented calibration gaps, not edits to the gold standard. Known-bad p018 ERROR pair unaffected. Local lint, schema check, ingest-check, pytest all pass.
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Unit 6 (
prompt-design), fifth of the MEDIUM batch, under preserve-by-default.Rubric (3 → 4): c1 unchanged · c2 = name the failure mode · c3 (NEW) = explain the mechanism · c4 = regime distinction (was c3).
Preserve-by-default decomposition
Faithful decomposition of the locked Opus values:
No realignments, no judgment-flips. p007/p011 labels updated for the 4-criterion shape.
Post-split distribution (21 pairs)
8 × 4-of-4 · 3 × 3-of-4 (p006, p007-differential, p008) · 2 × 2-of-4 (p010, p011) · 1 × 1-of-4 (p009) · 5 on-topic-all-missed · 2 off-topic.
Note
The known-bad p018 (emoji + structured markdown + slashed percentages — reproducible grader-payload ERROR per UNIT_6_GATE.md) is unaffected by the split; it remains a documented known-bad marker.
Local validation
lint_unit_markdown/ingest_units --check— cleanrun_regression_set --check— 21 pairs validpytest— 20/20Test plan
Opened as draft.
Generated by Claude Code