Common-sense Reasoning on ARC-e, ARC-c, BoolQ, HellaSwag, PIQA
74.7ARC-e AccuracySOLAR
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SOLARBase Model=Qwen2.5-0.5B-Instruct, Adaptation Method=SOLAR Level I2026.03 | 74.7 | 55.5 | 58.8 | 48.3 | 60.1 | 59.5 | |
| DnDBase Model=Qwen2.5-0.5B-Instruct2026.03 | 70.9 | 48.1 | 51.9 | 26.5 | 47.8 | 49 | |
| DOMBase Model=Qwen2.5-0.5B-Instruct2026.03 | 56.5 | 38.9 | 33.2 | 28.3 | 18.8 | 35.1 | |
| LoRABase Model=Qwen2.5-0.5B-Instruct2026.03 | 47.4 | 39.7 | 14.7 | 26.3 | 51.5 | 35.9 | |
| TTLBase Model=Qwen2.5-0.5B-Instruct2026.03 | 24.4 | 24.7 | 44.4 | 25.9 | 51.9 | 34.3 |