Multi-hop document retrieval on Musique (test)
0.83Recall@KIRCoT
Evaluation Results
| Method | Links | |
|---|---|---|
| IRCoTK=14.12025.03 | 0.83 | |
| SelfAskK=28.12025.03 | 0.765 | |
| L-RAGK=8.02025.03 | 0.764 | |
| Vanilla RAGK=8.02025.03 | 0.762 | |
| Vanilla RAGK=4.02025.03 | 0.695 | |
| ThinkGRModel Parameters=8B2026.05 | 0.6398 | |
| GritHopperModel Parameters=7B2026.05 | 0.6048 | |
| w/o RLModel Parameters=8B2026.05 | 0.534 | |
| w/o ThoughtModel Parameters=8B2026.05 | 0.5323 | |
| IterRetGenK=6.02025.03 | 0.523 | |
| L-RAGK=4.02025.03 | 0.52 | |
| R3-RAGModel Parameters=8B2026.05 | 0.517 | |
| RT-RAGModel Parameters=8B2026.05 | 0.513 | |
| IRCoTModel Parameters=70B2026.05 | 0.4996 | |
| Auto-RAGModel Parameters=7B2026.05 | 0.4063 | |
| ITER-RETGENModel Parameters=70B2026.05 | 0.3937 | |
| w/o SFTModel Parameters=8B2026.05 | 0.3598 | |
| SEALModel Parameters=406M2026.05 | 0.356 | |
| SelfaskModel Parameters=70B2026.05 | 0.3428 | |
| Contriever2026.05 | 0.3402 | |
| BGE-largeModel Parameters=326M2026.05 | 0.3339 | |
| BM252026.05 | 0.3157 | |
| HyDEK=6.02025.03 | 0.28 | |
| MDRModel Parameters=110M2026.05 | 0.2784 |