Multi-hop Question Answering on MuSiQue (EM, F1, Judge, Tokens (M), Cw)
51EMGRASP
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GRASPSetting=Extended Ctx.2026.05 | 51 | 64.3 | 72.5 | 8.87 | 46,707 | |
| GRASPSetting=Retrieval2026.05 | 50.5 | 65.1 | 76.5 | 15.89 | 12,616 | |
| GraphReaderSetting=Extended Ctx.2026.05 | 48 | 61.9 | 73 | 14.07 | 80,852 | |
| IRCoT + HippoRAG2Setting=Retrieval2026.05 | 47.8 | 60.4 | 67.3 | 31.49 | 26,655 | |
| GeARSetting=Retrieval2026.05 | 41.4 | 53 | 68.9 | 16.69 | 15,545 | |
| GraphReaderSetting=Extended Ctx., N=12026.05 | 40 | 53.8 | 63 | 9.44 | 73,719 | |
| ReadAgentSetting=Extended Ctx.2026.05 | 35.5 | 46.6 | 55.5 | 14.07 | 116,670 |