feat(extraction): preserve first-person negations in extraction prompt#3
Open
moralespanitz wants to merge 1 commit intomainfrom
Open
feat(extraction): preserve first-person negations in extraction prompt#3moralespanitz wants to merge 1 commit intomainfrom
moralespanitz wants to merge 1 commit intomainfrom
Conversation
BEAM CR diagnostic showed extraction was silently dropping user negative-tense statements like 'I have never used Flask-Login'. Without these, contradiction resolution is structurally unscoreable — only the positive side reaches storage. Validated with retrieval inspector: target negative statement now extracted verbatim and ranks #1 (sim=0.665) for the matching CR query.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Adds a
NEGATIVE STATEMENT PRESERVATIONrule to the extraction prompt so first-person negations like "I have never used X" are stored as explicit negative facts instead of being normalized to positives or dropped silently.Why
CR mini-slice (BEAM Sprint 2) showed extraction was silently dropping or inverting first-person negative assertions, which broke contradiction detection at the storage layer — AUDN had nothing to resolve because only the positive side ever reached the DB. n=6 baseline scored 0/6 on contradiction-resolution questions.
Validation
CR mini-slice run on commit
6b96a68— seedata/exp-cr-mini/in atomicmemory-benchmarks:"User has never written any Flask routes…"and"User has never integrated Flask-Login…"now persist as negative factsTest plan
/v1/memories/list)npx tsc --noEmitpassesnpm testpasses (no behavioral changes to extraction call flow, only prompt text)fallow --no-cacheclean