Web-based Question Answering on BrowseComp-plus
78.41AccuracyGPT5/MixedbreadSearch
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT5/MixedbreadSearchLLM=GPT5, Retriever=MixedbreadSearch, Rank=1st2026.01 | 78.41 | 48.85 | 44.67 | — | |
| GPT5/Q3-E8BLLM=GPT5, Retriever=Q3-E8B, Rank=3rd2026.01 | 71.69 | 78.98 | 21.74 | — | |
| TDR-30B-A3B/FABLELLM=TDR-30B-A3B, Retriever=FABLE, Selection Strategy=bi-path selection2026.01 | 66.6 | 76.6 | 21.74 | — | |
| o3/Q3-E8BLLM=o3, Retriever=Q3-E8B, Rank=4rd2026.01 | 65.9 | 73.24 | 23.97 | — | |
| TDR-30B-A3B/FABLE(ln)LLM=TDR-30B-A3B, Retriever=FABLE(ln), Selection Strategy=LLM-based node selection2026.01 | 64.92 | 68.07 | 46.99 | — | |
| GPT5/BM25LLM=GPT5, Retriever=BM25, Rank=5rd2026.01 | 57.59 | 61.7 | 23.23 | — | |
| TDR-30B-A3B/Q3-E8BLLM=TDR-30B-A3B, Retriever=Q3-E8B, Rank=11th2026.01 | 44.46 | 62.32 | 30.37 | — | |
| Sonnet 42025.07 | 37.35 | 47.33 | 9.03 | — | |
| Opus 42025.07 | 36.75 | 50.84 | 10.24 | — | |
| kimi-k2-0711-preview2025.07 | 35.42 | 38.38 | 11.22 | — | |
| oss-20b-highModel Size=20B2025.07 | 35.06 | 49.29 | 23.87 | — | |
| Gemini-2.5-Flash2025.07 | 34.58 | 40.19 | 9.77 | — | |
| oss-20b-mediumModel Size=20B2025.07 | 30.48 | 41.31 | 13.64 | — | |
| Gemini-2.5-Pro2025.07 | 29.52 | 35.31 | 6.04 | — | |
| oss-120b-lowModel Size=120B2025.07 | 25.54 | 22.5 | 2.21 | — | |
| FRUGALRAG-7B + Qwen2.5-7B-InstModel Size=7B, Training Data=2Wiki2025.07 | 21.53 | 22.93 | 10.96 | — | |
| FRUGALRAG-7B + Qwen2.5-7B-InstModel Size=7B, Training Data=MuSiQue2025.07 | 21.14 | 23.73 | 8.39 | — | |
| FRUGALRAG-7B + Qwen2.5-7B-InstModel Size=7B, Training Data=HotPotQA2025.07 | 20.46 | 23.57 | 7.95 | — | |
| DeepSeek-R1-0528Model Size=600B2025.07 | 16.39 | 16.32 | 2.72 | — | |
| oss-20b-lowModel Size=20B2025.07 | 14.1 | 17.37 | 1.87 | — | |
| Search-R1-32BModel Size=32B2025.07 | 11.08 | 10.17 | 1.69 | — | |
| Qwen3-32BModel Size=32B2025.07 | 10.72 | 7.8 | 0.94 | — | |
| Adaptive-RAG (Extended)Accuracy Goal=90%2026.05 | — | — | — | 1.1 | |
| BRANEAccuracy Goal=100%2026.05 | — | — | — | 81.7 | |
| BRANEAccuracy Goal=95%2026.05 | — | — | — | 88 | |
| BRANEAccuracy Goal=90%2026.05 | — | — | — | 91 | |
| CARROT-KNN (Extended)Accuracy Goal=95%2026.05 | — | — | — | 6.6 | |
| CARROT-KNN (Extended)Accuracy Goal=90%2026.05 | — | — | — | 29.7 | |
| CARROT-RoBERTa (Extended)Accuracy Goal=100%2026.05 | — | — | — | 0 | |
| CARROT-RoBERTa (Extended)Accuracy Goal=95%2026.05 | — | — | — | 26.9 | |
| CARROT-RoBERTa (Extended)Accuracy Goal=90%2026.05 | — | — | — | 41.9 | |
| Most Accurate Static (Murakkab)Accuracy Goal=100%2026.05 | — | — | — | 0 | |
| Most Accurate Static (Murakkab)Accuracy Goal=95%2026.05 | — | — | — | 0 | |
| Most Accurate Static (Murakkab)Accuracy Goal=90%2026.05 | — | — | — | 0 |