Reasoning on ARB OOD
88.7AccuracyGPT-4.5
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-4.52026.06 | 88.7 | |
| IGA (Ours)2026.06 | 79.4 | |
| Pareto-GS2026.06 | 73.2 | |
| SAND-Mask2026.06 | 71.4 | |
| PCGrad2026.06 | 70.5 | |
| V-REx2026.06 | 69.8 | |
| IRM2026.06 | 69.1 | |
| ERM-SFT2026.06 | 65.1 | |
| CoT-Distill2026.06 | 64.2 | |
| LoRA-SFT2026.06 | 63.9 |