Multi-hop Question Answering on HotpotQA (F1, EM)
61.63F1 ScoreSEARCH-R
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SEARCH-RBackbone=Llama-3.1-8B+DeepSeek-V32026.04 | 61.63 | 46 | |
| SEARCH-RBackbone=Llama-3.1-8B+GPT-4o-mini2026.04 | 60.06 | 44 | |
| ChainRAGBackbone=GPT-4o-mini2026.04 | 59.13 | 44.5 | |
| Iter-RetGenBackbone=GPT-4o-mini2026.04 | 57.77 | 42 | |
| NativeRAGBackbone=GPT-4o-mini2026.04 | 56.92 | 42 | |
| HippoRAGBackbone=GPT-4o-mini2026.04 | 56.12 | 40 | |
| ChainRAGBackbone=DeepSeek-V32026.04 | 53.99 | 41 | |
| SEARCH-RBackbone=Llama-3.1-8B2026.04 | 49.09 | 36 | |
| NativeRAGBackbone=Llama-3.1-8B2026.04 | 47 | 33.5 | |
| NativeRAGBackbone=DeepSeek-V32026.04 | 46.19 | 32.5 | |
| ChainRAGBackbone=Llama-3.1-8B2026.04 | 45.21 | 34 | |
| Search-R1Backbone=Llama-3.1-8B2026.04 | 43.52 | 30.5 |