General Reasoning on BGQA, CRUX, STGQA, TableBench, MMLU STEM Suite
60.4BGQA AccuracyCASPO+CaT
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| CASPO+CaTBase Model=Qwen2.5-7B-Instruct2026.05 | 60.4 | 66 | 98.5 | 48 | 50.4 | 64.7 | |
| CASPOBase Model=Qwen2.5-7B-Instruct2026.05 | 57.5 | 62.9 | 93.8 | 45.7 | 48 | 61.6 | |
| DPO-VPBase Model=Qwen2.5-7B-Instruct2026.05 | 56.8 | 62.1 | 93.3 | 45.2 | 47.4 | 61 | |
| CASPO+CaTBase Model=Qwen2.5-Math-7B2026.05 | 56.2 | 58.3 | 96.1 | 43.6 | 46.2 | 60.1 | |
| PURE-VRBase Model=Qwen2.5-7B-Instruct2026.05 | 56 | 61.4 | 92.8 | 44.7 | 47 | 60.4 | |
| Simple-RL-ZeroBase Model=Qwen2.5-7B-Instruct2026.05 | 55.5 | 60.9 | 92.5 | 44.4 | 46.7 | 60 | |
| GRPOBase Model=Qwen2.5-7B-Instruct2026.05 | 54.5 | 59.9 | 92.1 | 44 | 46.2 | 59.3 | |
| CASPOBase Model=Qwen2.5-Math-7B2026.05 | 53.5 | 55.5 | 91.5 | 41.5 | 44 | 57.2 | |
| Qwen2.5-7B-InstructBase Model=Qwen2.5-7B-Instruct2026.05 | 53 | 58.1 | 91.3 | 43.2 | 45.2 | 58.2 | |
| DPO-VPBase Model=Qwen2.5-Math-7B2026.05 | 52.5 | 54.5 | 91 | 41 | 43.5 | 56.5 | |
| PURE-VRBase Model=Qwen2.5-Math-7B2026.05 | 52 | 54 | 90.5 | 40.5 | 43 | 56 | |
| Simple-RL-ZeroBase Model=Qwen2.5-Math-7B2026.05 | 51.5 | 53.5 | 90 | 40 | 42.5 | 55.5 | |
| GRPOBase Model=Qwen2.5-Math-7B2026.05 | 50.5 | 53 | 89.5 | 39 | 42 | 54.8 | |
| Qwen2.5-Math-7BBase Model=Qwen2.5-Math-7B2026.05 | 48 | 50 | 88 | 38 | 40 | 52.8 | |
| CASPO+CaTBase Model=Llama-3.1-8B-Instruct2026.05 | 45.7 | 51 | 88.8 | 39.4 | 40.4 | 53.1 | |
| CASPOBase Model=Llama-3.1-8B-Instruct2026.05 | 43.5 | 48.5 | 84.5 | 37.5 | 38.5 | 50.5 | |
| DPO-VPBase Model=Llama-3.1-8B-Instruct2026.05 | 43 | 48 | 84.2 | 37 | 38 | 50 | |
| PURE-VRBase Model=Llama-3.1-8B-Instruct2026.05 | 42.5 | 47.5 | 83.8 | 36.5 | 37.5 | 49.6 | |
| Simple-RL-ZeroBase Model=Llama-3.1-8B-Instruct2026.05 | 42 | 47 | 83.5 | 36.2 | 37.2 | 49.2 | |
| GRPOBase Model=Llama-3.1-8B-Instruct2026.05 | 41.5 | 46.5 | 83 | 35.8 | 37 | 48.8 | |
| Llama-3.1-8B-InstructBase Model=Llama-3.1-8B-Instruct2026.05 | 40 | 45 | 82 | 35 | 36 | 47.6 |