Long-context Retrieval on BRIGHT StackExchange
62.3Biology ScoreClaude Sonnet 4.5
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Claude Sonnet 4.5Category=Ours (AgenticRAG), Tools=search, find, open, summ.2026.05 | 62.3 | 60 | 58.7 | 67.9 | 55 | 34.1 | 51.7 | 7.1 | 49.6 | |
| Claude Sonnet 4.5Category=Agentic, Strategy=search/find/open2026.05 | 62.3 | 60 | 58.7 | 67.9 | 55 | 34.1 | 51.7 | 7.1 | 49.6 | |
| GPT-5-miniCategory=Ours (AgenticRAG), Tools=search, find, open, summ.2026.05 | 61.7 | 48.1 | 41.4 | 65.3 | 39.4 | 40.6 | 46.6 | 4.8 | 43.5 | |
| GPT-5-miniCategory=Agentic, Strategy=search/find/open2026.05 | 61.7 | 48.1 | 41.4 | 65.3 | 39.4 | 40.6 | 46.6 | 4.8 | 43.5 | |
| QwenCategory=Open-source Emb.2026.05 | 39.2 | 36.1 | 25.7 | 42.3 | 21.3 | 23.5 | 33.1 | 1.3 | 27.8 | |
| QwenCategory=Open-sourced embedding models2026.05 | 39.2 | 36.1 | 25.7 | 42.3 | 21.3 | 23.5 | 33.1 | 1.3 | 27.8 | |
| GPT-4Base Retriever=SBERT, Reasoning Enhanced=true2026.05 | 37.7 | 35.3 | 19.9 | 18.3 | 12.4 | 11.5 | 22.6 | — | 22.5 | |
| GritLMCategory=Open-sourced embedding models2026.05 | 37.5 | 40.3 | 25.7 | 34.4 | 17.8 | 20.1 | 32.4 | 0 | 26 | |
| ReDIBase Retriever=SBERT, Reasoning Enhanced=true2026.05 | 36.2 | 32.8 | 22.8 | 20.8 | 10.9 | 16.2 | 22.2 | — | 23.1 | |
| DeepSeek-R1Base Retriever=SBERT, Reasoning Enhanced=true2026.05 | 35.6 | 34.8 | 16 | 15.3 | 8.9 | 15 | 19.9 | — | 20.8 | |
| Claude-3-OpusBase Retriever=SBERT, Reasoning Enhanced=true2026.05 | 34.8 | 31.6 | 21.8 | 15.8 | 8.9 | 15.8 | 16.6 | — | 20.8 | |
| VoyageCategory=Proprietary Emb.2026.05 | 34.4 | 35.4 | 26.7 | 41.6 | 12.9 | 12.8 | 31.1 | 1.3 | 24.5 | |
| VoyageCategory=Proprietary embedding models2026.05 | 34.4 | 35.4 | 26.7 | 41.6 | 12.9 | 12.8 | 31.1 | 1.3 | 24.5 | |
| OpenAICategory=Proprietary embedding models2026.05 | 32.1 | 31.4 | 23.8 | 34.2 | 11.9 | 10.7 | 26.3 | 0 | 21.3 | |
| CohereCategory=Proprietary embedding models2026.05 | 31.5 | 34.5 | 18.9 | 20.5 | 9.9 | 15.8 | 15.2 | 0.8 | 18.4 | |
| GoogleCategory=Proprietary embedding models2026.05 | 30.9 | 38 | 21.9 | 30.7 | 12.9 | 19.2 | 25.7 | 0.3 | 22.4 | |
| SFRCategory=Open-sourced embedding models2026.05 | 30.3 | 37 | 24.3 | 47.7 | 17.3 | 14.5 | 35 | 2 | 26 | |
| E5Category=Open-sourced embedding models2026.05 | 29.9 | 36.3 | 26.2 | 46.7 | 17.3 | 14.5 | 32.2 | 1.1 | 25.5 | |
| ReDICategory=Reasoning Enhanced2026.05 | 28.4 | 22.4 | 21.2 | 32 | 19.8 | 36.3 | 21.7 | — | 26 | |
| ReDIBase Retriever=BM25, Reasoning Enhanced=true2026.05 | 28.4 | 22.4 | 21.2 | 32 | 19.8 | 36.3 | 21.7 | — | 26 | |
| Claude-3-OpusBase Retriever=BM25, Reasoning Enhanced=true2026.05 | 26.8 | 13.5 | 13.4 | 28.2 | 7.9 | 28.2 | 11.8 | — | 18.5 | |
| GPT-4Base Retriever=BM25, Reasoning Enhanced=true2026.05 | 26.8 | 15.8 | 10.2 | 30.7 | 5.9 | 26.5 | 9.7 | — | 17.9 | |
| DeepSeek-R1Base Retriever=BM25, Reasoning Enhanced=true2026.05 | 26.8 | 20 | 14.4 | 30.2 | 14.9 | 33.3 | 10.6 | — | 21.5 | |
| SBERTCategory=Open-sourced embedding models2026.05 | 25.6 | 34.1 | 18.9 | 15.8 | 10.9 | 15 | 18 | 1.2 | 17.4 | |
| Inst-LCategory=Open-sourced embedding models2026.05 | 24.6 | 29.9 | 13.1 | 20.3 | 12.9 | 15 | 25.4 | 3.9 | 18.1 | |
| Inst-XLCategory=Open-sourced embedding models2026.05 | 21.5 | 31 | 13.1 | 20.5 | 13.9 | 15 | 20.1 | 6 | 17.6 | |
| BGECategory=Open-sourced embedding models2026.05 | 16.4 | 27.7 | 20.9 | 11.6 | 10.9 | 13.3 | 16.9 | 0.4 | 14.8 | |
| BM25Category=Sparse2026.05 | 10.7 | 15.4 | 10.7 | 8.4 | 7.4 | 22.2 | 10.7 | 5.4 | 11.4 | |
| BM25Category=Sparse models2026.05 | 10.7 | 15.4 | 10.7 | 8.4 | 7.4 | 22.2 | 10.7 | 5.4 | 11.4 |