Reasoning on ARC-C, BBH, GPQA, CALM, and KOR-BENCH
93.9ARC-C ScoreOmniThought-0528
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| OmniThought-0528Backbone=Qwen3-8B-Base, Size=365k, Eff.=+0.0192025.12 | 93.9 | 86.8 | 51 | 33.4 | 65 | 66 | |
| SYNTHETIC-2-SFTBackbone=Qwen2.5-7B-Base, Size=105k, Eff.=+0.1542025.12 | 93.2 | 81.1 | 31.8 | 29.5 | 58.8 | 58.9 | |
| MiroMind-M1-SFTBackbone=Qwen3-8B-Base, Size=719k, Eff.=+0.0072025.12 | 92.9 | 86.5 | 47.5 | 40.2 | 62.1 | 65.8 | |
| SYNTHETIC-2-SFTBackbone=Qwen3-8B-Base, Size=105k, Eff.=+0.0692025.12 | 92.9 | 86.6 | 55.6 | 34 | 64.2 | 66.6 | |
| Light-R1-SFTBackbone=Qwen3-8B-Base, Size=79k, Eff.=+0.0692025.12 | 92.2 | 84.5 | 46.5 | 34.2 | 60.4 | 63.6 | |
| ODA-Mixture-500kBackbone=Qwen3-8B-Base, Size=506k, Eff.=+0.0382025.12 | 92.2 | 88 | 48 | 54.8 | 65.8 | 69.7 | |
| ODA-Mixture-100kBackbone=Qwen3-8B-Base, Size=101k, Eff.=+0.0872025.12 | 91.5 | 82.4 | 45 | 45 | 59.4 | 64.7 | |
| OpenThoughts-114kBackbone=Qwen3-8B-Base, Size=114k, Eff.=+0.0362025.12 | 91.2 | 82.3 | 42.4 | 43.1 | 57.8 | 63.3 | |
| OmniThought-0528Backbone=Qwen2.5-7B-Base, Size=365k, Eff.=+0.0402025.12 | 90.5 | 76.8 | 30.8 | 29.1 | 58.9 | 57.2 | |
| ODA-Mixture-500kBackbone=Qwen2.5-7B-Base, Size=506k, Eff.=+0.0332025.12 | 88.1 | 82.7 | 31.3 | 35 | 58.8 | 59.6 | |
| Orca-agentinstructBackbone=Qwen2.5-7B-Base, Size=1.2M, Eff.=+0.0102025.12 | 86.8 | 70.2 | 31.3 | 38.6 | 48.5 | 55.1 | |
| ODA-Mixture-100kBackbone=Qwen2.5-7B-Base, Size=101k, Eff.=+0.0872025.12 | 84.1 | 68.1 | 22.7 | 33.2 | 49.4 | 51.5 | |
| OpenThoughts3Backbone=Qwen2.5-7B-Base, Size=1.2M, Eff.=+0.0102025.12 | 82.7 | 70.7 | 28.8 | 32.9 | 56.2 | 54.3 | |
| OpenO1-SFTBackbone=Qwen2.5-7B-Base, Size=78k, Eff.=+0.1462025.12 | 80.7 | 72.5 | 33.8 | 36 | 47.2 | 54 | |
| Qwen3-8B-BaseBackbone=Qwen3-8B-Base2025.12 | 37.3 | 78.1 | 41.9 | 48.9 | 46.6 | 50.6 | |
| Qwen2.5-7B-BaseBackbone=Qwen2.5-7B-Base2025.12 | 36.6 | 69.5 | 34.9 | 39.2 | 33.3 | 42.7 |