Multi-hop Retrieval on Morehopqa (test)
80.5RecallThinkGR
Evaluation Results
| Method | Links | |
|---|---|---|
| ThinkGRModel Parameters=8B2026.05 | 80.5 | |
| w/o ThoughtModel Parameters=8B2026.05 | 78 | |
| GritHopperModel Parameters=7B2026.05 | 74.82 | |
| w/o RLModel Parameters=8B2026.05 | 73.84 | |
| R3-RAGModel Parameters=8B2026.05 | 70.44 | |
| RT-RAGModel Parameters=8B2026.05 | 66.95 | |
| IRCoTModel Parameters=70B2026.05 | 66.82 | |
| ITER-RETGENModel Parameters=70B2026.05 | 60.73 | |
| Auto-RAGModel Parameters=7B2026.05 | 59.48 | |
| SelfaskModel Parameters=70B2026.05 | 57.6 | |
| w/o SFTModel Parameters=8B2026.05 | 55.14 | |
| MDRModel Parameters=110M2026.05 | 49.6 | |
| BGE-largeModel Parameters=326M2026.05 | 47.58 | |
| SEALModel Parameters=406M2026.05 | 47.27 | |
| Contriever2026.05 | 45.04 | |
| BM252026.05 | 43.74 |