Release v0.2.0: grounding heatmap + per-sentence explanations #18
Merged
Conversation
…tions Compute source support for every summary sentence from the shared NLI matrix instead of only the gated-suspicious ones, so grounded sentences highlight their supporting source too. An entailment floor means a fabricated sentence surfaces no supporters rather than lighting up the whole document. The gated Inseq score stays the fusion feature, so the trained model is untouched. - types: Evidence gains source_support (source_id, entailment) for the heatmap - support: returns graded (id, score) pairs above a support floor - pipeline: run support attribution ungated over all sentences for the heatmap - app: clickable graded source panel; span-level highlighting of flagged tokens; a "why this verdict" panel (signal scores, best supporting source or a no-support note, failed claims); source-to-summary linking on click
feat(ui): generator-agnostic grounding heatmap + per-sentence explanations
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Promotes
dev→mainfor the v0.2.0 release.Highlights
summary sentence highlights its supporting source sentences (graded bins:
strong support ≥0.80 / support / neutral); span-level highlighting of
fabricated tokens in the summary.
"no support" contrastive evidence, and failed-claim listing; bidirectional
source↔summary linking.
support_attribution(NLI matrix → conc/loo +support pairs) run ungated over all sentences for the heatmap, while the gated
Inseq scalar remains the fusion feature (trained model unchanged).