Multi-hop Question Answering on HotpotQA fullwiki (val)
58.3Exact Match (EM)SDP
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SDPLLM=GPT-4o2026.05 | 58.3 | 67.2 | — | |
| PRISMLLM=GPT-4o2026.05 | 54.2 | 67 | — | |
| IRCoTLLM=GPT-32026.05 | 49.3 | 60.7 | — | |
| DynaKRAGBackbone=GPT-4o-mini2026.07 | 49.22 | 62.18 | 1,588.5 | |
| DynaKRAGBackbone=Qwen2.5-7B2026.07 | 47.32 | 59.98 | 2,373.1 | |
| S2G-RAGBackbone=GPT-4o-mini2026.07 | 44.98 | 60.95 | 1,766 | |
| CoRAG-s3Backbone=GPT-4o-mini2026.07 | 44.96 | 61.08 | 2,569.7 | |
| IRCoTBackbone=GPT-4o-mini2026.07 | 44.85 | 60.83 | 1,545.7 | |
| S2G-RAGBackbone=Qwen2.5-7B2026.07 | 44.25 | 57.1 | 2,807.3 | |
| CRAGBackbone=GPT-4o-mini2026.07 | 44.06 | 59.51 | 1,355.4 | |
| Self-Ask+SearchBackbone=GPT-4o-mini2026.07 | 44.04 | 60.23 | 3,432.1 | |
| IRCoTBackbone=Qwen2.5-7B2026.07 | 43.39 | 56.17 | 1,723.5 | |
| CoRAG-s3Backbone=Qwen2.5-7B2026.07 | 42.82 | 55.73 | 2,814.3 | |
| CRAGBackbone=Qwen2.5-7B2026.07 | 41.81 | 54.45 | 1,720.1 | |
| Fixed-KBackbone=GPT-4o-mini2026.07 | 41.7 | 57.81 | 1,745.1 | |
| Self-Ask+SearchBackbone=Qwen2.5-7B2026.07 | 40.43 | 52.73 | 3,302 | |
| Fixed-KBackbone=Qwen2.5-7B2026.07 | 39.58 | 51.77 | 1,892.3 | |
| SETR-CoT & IRILLM=GPT-4o2026.05 | 39.2 | 40.5 | — | |
| PAR2-RAGBackbone=GPT-4o-mini2026.07 | 39.18 | 55.02 | 2,776 | |
| Adaptive-RAGBackbone=Qwen2.5-7B2026.07 | 38.62 | 50.42 | 1,596.4 | |
| PAR2-RAGBackbone=Qwen2.5-7B2026.07 | 37.79 | 49.52 | 3,053.6 | |
| DynaKRAGBackbone=Llama-3.1-8B2026.07 | 37.42 | 48.76 | 1,868.6 | |
| CRAGBackbone=Llama-3.1-8B2026.07 | 35.64 | 47.7 | 968.8 | |
| S2G-RAGBackbone=Llama-3.1-8B2026.07 | 35.17 | 47.61 | 2,843 | |
| RankZephyrLLM=GPT-4o2026.05 | 34.7 | 35 | — | |
| RankGPTLLM=GPT-4o2026.05 | 34.6 | 35.3 | — | |
| Adaptive-RAGBackbone=GPT-4o-mini2026.07 | 34.19 | 47.83 | 1,120.4 | |
| RankZephyr + CoTLLM=GPT-4o2026.05 | 34 | 34.4 | — | |
| CoRAG-s3Backbone=Llama-3.1-8B2026.07 | 32.84 | 44.99 | 2,576.3 | |
| IRCoTBackbone=Llama-3.1-8B2026.07 | 32.52 | 44.66 | 1,657.1 | |
| Adaptive-RAGBackbone=Llama-3.1-8B2026.07 | 32.15 | 43.95 | 1,197.2 | |
| Self-Ask+SearchBackbone=Llama-3.1-8B2026.07 | 32.11 | 43.54 | 3,525.1 | |
| PAR2-RAGBackbone=Llama-3.1-8B2026.07 | 31.56 | 43.21 | 3,035.5 | |
| Fixed-KBackbone=Llama-3.1-8B2026.07 | 30.29 | 43.11 | 1,799.4 |