Reasoning on ARB ID
91.2Accuracy (ARB ID)GPT-4.5
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-4.52026.06 | 91.2 | |
| IGA (Ours)2026.06 | 82.1 | |
| Pareto-GS2026.06 | 79.6 | |
| SAND-Mask2026.06 | 78.9 | |
| PCGrad2026.06 | 78.1 | |
| V-REx2026.06 | 77.9 | |
| IRM2026.06 | 77.4 | |
| ERM-SFT2026.06 | 75.3 | |
| CoT-Distill2026.06 | 74.1 | |
| LoRA-SFT2026.06 | 73.8 |