Spaces:
Sleeping
Sleeping
Commit History
-
tune(rag): calibrate relevance filter for raw logit scores
- GAP_THRESHOLD: 0.3 → 5.0 (raw cross-encoder logits, not normalized) - SCORE_FLOOR: add -9.0 threshold for off-topic detection - Improved LLM judge prompt with archive scope context - Benchmark comparison mode in benchmark_filter.py - Results: recall 0.95, latency P50 777ms (vs ~10s before) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>0110ecb -
feat(eval): add benchmark script to collect Gemini filter ground truth
Runs 30 eval queries through retriever + Gemini generation and records which story UIDs survive the citation filter, establishing ground truth for future filter replacement comparisons. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>e623f4f