asmhatre/spring-ai
1
0
Fork 0
Code Activity

Add rerank module: bi-encoder retrieval plus an ONNX Runtime cross-encoder reranker behind a Spring AI DocumentPostProcessor; hit@k and MRR by pool size and question kind, latency by candidates, batch size, threads and int8, and five reproduced traps. Local MiniLM models, no database or API key.

Browse Source 
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01Cfo2K4u4rF18QsPXPFnytf
This commit is contained in:
Claude
2026-10-24 18:00:05 +00:00
parent 2ccf1d8d19
commit a429aa40f6
34 changed files with 1617 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rerank/output/03-quality-by-pool-size.txt
+16
View File
@@ -0,0 +1,16 @@
# Rerank the bi-encoder's top N (60 questions, 94 passages, fp32 cross-encoder)
ranking hit@1 hit@3 hit@5 MRR@10
bi-encoder alone 41/60 51/60 55/60 0.782
bi-encoder top 5 + cross-encoder 48/60 55/60 55/60 0.863
bi-encoder top 10 + cross-encoder 49/60 54/60 56/60 0.868
bi-encoder top 20 + cross-encoder 49/60 55/60 56/60 0.870
bi-encoder top 50 + cross-encoder 49/60 56/60 56/60 0.869
bi-encoder top 94 + cross-encoder 49/60 56/60 56/60 0.869
recall of the first stage (is the gold passage inside the pool at all?):
top 5 : 55/60
top 10 : 58/60
top 20 : 59/60
top 50 : 60/60
top 94 : 60/60