perf: keyword-first hybrid fusion, lighter query-side embedding - #49
Merged
Merged
Conversation
…ng the model, hybrid >= keyword CI floor
…sion only on package files (fetched 96/105)
…e lists five configurations
Merged
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Hybrid search is no longer worse than keyword search: on package files, weighted reciprocal rank fusion (90% keyword, 10% embedding) replaces the min-max score mix, which let the embedding outvote exact API-name matches. The score mix is kept when upstream docs are indexed (fetched 96/105). Benchmark, package files: hybrid 60 to 64 of 105, against keyword-only 59; held-out 8/18, which ties keyword-only, so the gain on unseen questions is not proven yet. A query reads only its own tokens' rows from the model file: median peak RSS 48 MB to 18 MB, latency 42 ms to 18 ms. CI fails if hybrid scores below keyword-only. Opus review: GO.