Abstract Reasoning on ARC Community
63.04AccuracyQwen3 + DIARC
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3 + DIARC#Params=3.6B2026.06 | 63.04 | |
| Qwen3 + SFT only#Params=3.6B2026.06 | 58.7 | |
| Minitron + DIARC#Params=6.9B2026.06 | 43.48 | |
| Minitron + SFT only#Params=6.9B2026.06 | 34.78 | |
| Llama-3.2 + DIARC#Params=2.6B2026.06 | 30.43 | |
| Llama-3.2 + SFT only#Params=2.6B2026.06 | 23.91 | |
| Minitron#Params=8B2026.06 | 6.52 | |
| Llama-3.2#Params=3B2026.06 | 2.17 | |
| Qwen3#Params=4B2026.06 | 0 |