Dev - #4
Merged
Merged
Conversation
Validating draft hooks as verbatim source substrings was brittle: faithful paraphrases were discarded, so the first full Cerebras eval produced groundedness=0.0 with empty verdicts for most qualified companies — an artifact of the matching layer, not of grounding. Source-text grounding belongs at extraction; the draft should ground by selecting facts and the verify gate should judge the body's faithfulness (ADR-0014). - draft node grounds by fact-selection (model returns fact ids; hooks_used grounded by construction; no substring/fuzzy hook matching). - verify node = structural hook check + one body-faithfulness judge (judge_body) over the draft body vs the selected facts. groundedness_score redefined = faithful_body_claims / total_body_claims. Replaces judge_faithfulness; failure reasons -> unsupported/overreach/structural/ judge-error. - VerificationResult/ClaimVerdict now describe body claims; tier_breakdown counts grounding hooks. - new `redraft` eval command: re-run draft+verify only, reusing cached research AND frozen qualification verdicts. Re-ran draft+verify only on cerebras/gpt-oss-120b (2026-06-14); qualification frozen (TP=10 FP=3 TN=4 FN=0, F1 0.870). Draft-gate 11/13=0.846, mean groundedness 0.936, own_site live re-verifiability 0.90 (45/50). docs/evals.md records the real numbers + a Known Limitations section (3 qualification false positives; the industry-gating fix is deferred future work, not tuned on this set). Docs, ADR-0014, CHANGELOG, roadmap updated; pytest green (no network), mkdocs build --strict green. Note: lands on top of the in-flight 0.6.0 (lean research depth) and 0.7.0 (Cerebras provider) working-tree changes.
…scored component (ADR-0015); temperature 0
…kens; temperature 0
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.
No description provided.