End-to-End Retrieval-Augmented Generation on RedQA Legal
45.82E2E ScoreRARE (RAG)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RARE (RAG)Generator=GPT-5 Mini, Judge=GPT-5, Retrieval Context=Average of 9 retrievers2026.04 | 45.82 | — | 19.7 | 34.84 | 10.98 | |
| GPT-5 Mini (Param-only)Generator=GPT-5 Mini, Retrieval Context=None2026.04 | — | 26.12 | — | — | — |