The RAG Index / Reranking / #137

hakari-bench/hakari-bench

by hakari-bench · Reranking · updated 11d ago

⚖️ HAKARI-Bench is a lightweight IR benchmark that rebuilds retrieval tasks as small Nano-sets, making model selection, regression checks, quantization, truncation, and reranking comparisons practical under the same conditions.

46
momentum
31
stars
1
forks
#137
rank
nlp
View on GitHub →