Question Answering on OBQA (Accuracy, Avg, ΔScore)
88.8AccuracyETN
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ETN2026.04 | 88.8 | 97.15 | — | 94.57 | |
| DMM2026.04 | 87.2 | 95.3 | — | 93.46 | |
| MAP_CE2026.04 | 87.07 | 96.94 | — | — | |
| MCD2026.04 | 86.4 | 97.18 | — | — | |
| LL2026.04 | 86.07 | 51.27 | — | — | |
| DeepEns2026.04 | 83.47 | 94.51 | — | — | |
| R1-OracleBackbone=Qwen2.5-3B-Instruct, Supervision Paradigm=100% GT2026.03 | 83.11 | 79.12 | 0 | — | |
| MemRewardBackbone=Qwen2.5-3B-Instruct, Supervision Paradigm=20% GT + GNN2026.03 | 81.78 | 77.02 | -2.1 | — | |
| IB-EDL2026.04 | 81.6 | 94.33 | — | 83.1 | |
| MAP_EDL2026.04 | 80.27 | 85.62 | — | 76.69 | |
| R1-pBackbone=Qwen2.5-3B-Instruct, Supervision Paradigm=20% GT2026.03 | 78.44 | 75.67 | -3.45 | — | |
| R1-OracleBackbone=Qwen2.5-1.5B-Instruct, Supervision Paradigm=100% GT2026.03 | 74 | 70.47 | 0 | — | |
| MemRewardBackbone=Qwen2.5-1.5B-Instruct, Supervision Paradigm=20% GT + GNN2026.03 | 70 | 68.1 | -2.37 | — | |
| R1-pBackbone=Qwen2.5-1.5B-Instruct, Supervision Paradigm=20% GT2026.03 | 68.67 | 62.72 | -7.75 | — |