Aggregate Benchmark Performance on Math500, GSM8K, MMLU-s, GPQA, AIME Suite
42.99Average P@1ICT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ICTModel Size=Qwen2.5-7B2026.06 | 42.99 | 58.37 | 50.68 | |
| STAPOModel Size=Qwen2.5-7B2026.06 | 41.55 | 55.52 | 48.54 | |
| GRPOModel Size=Qwen2.5-7B2026.06 | 39.4 | 53.16 | 46.28 | |
| BaseModel Size=Qwen2.5-7B2026.06 | 37.78 | 55.25 | 46.52 | |
| 20-EntropyModel Size=Qwen2.5-7B2026.06 | 37.53 | 53.29 | 45.41 | |
| ICTModel Size=Qwen2.5-1.5B2026.06 | 21.93 | 36.53 | 29.23 | |
| STAPOModel Size=Qwen2.5-1.5B2026.06 | 19.89 | 31.52 | 25.71 | |
| 20-EntropyModel Size=Qwen2.5-1.5B2026.06 | 18.66 | 31.02 | 24.84 | |
| GRPOModel Size=Qwen2.5-1.5B2026.06 | 18.22 | 31.87 | 25.05 | |
| BaseModel Size=Qwen2.5-1.5B2026.06 | 16.41 | 31.8 | 24.11 | |
| ICTModel Size=Qwen2.5-0.5B2026.06 | 10.95 | 21.75 | 16.35 | |
| GRPOModel Size=Qwen2.5-0.5B2026.06 | 10.52 | 18.97 | 14.75 | |
| STAPOModel Size=Qwen2.5-0.5B2026.06 | 9.54 | 18.29 | 13.92 | |
| 20-EntropyModel Size=Qwen2.5-0.5B2026.06 | 8.26 | 16.37 | 12.32 | |
| BaseModel Size=Qwen2.5-0.5B2026.06 | 5.65 | 14.48 | 10.07 |