Reasoning on Average Multi-domain
10,572Thinking Tokens CountDRE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DREBase Model=Qwen3-8B2026.06 | 10,572 | 63.86 | 6.04 | |
| DREBase Model=Qwen3-30B-A3B-Thinking-25072026.06 | 10,979 | 79.78 | 7.27 | |
| DREBase Model=Qwen3-4B-Thinking-25072026.06 | 12,545 | 72.25 | 5.76 | |
| GRPO + LPBase Model=Qwen3-8B2026.06 | 13,030 | 62.58 | 4.8 | |
| GRPOBase Model=Qwen3-30B-A3B-Thinking-25072026.06 | 13,238 | 78.66 | 5.94 | |
| GRPOBase Model=Qwen3-8B2026.06 | 13,365 | 63.15 | 4.73 | |
| Raw ModelBase Model=Qwen3-8B2026.06 | 13,609 | 63.62 | 4.67 | |
| GRPO + LPBase Model=Qwen3-30B-A3B-Thinking-25072026.06 | 13,710 | 79.12 | 5.77 | |
| Raw ModelBase Model=Qwen3-30B-A3B-Thinking-25072026.06 | 14,572 | 80.76 | 5.54 | |
| S-GRPOBase Model=Qwen3-8B2026.06 | 14,777 | 62.23 | 4.21 | |
| S-GRPOBase Model=Qwen3-30B-A3B-Thinking-25072026.06 | 15,897 | 78.89 | 4.96 | |
| GRPOBase Model=Qwen3-4B-Thinking-25072026.06 | 16,739 | 71.51 | 4.27 | |
| GRPO + LPBase Model=Qwen3-4B-Thinking-25072026.06 | 17,051 | 70.47 | 4.13 | |
| Raw ModelBase Model=Qwen3-4B-Thinking-25072026.06 | 17,407 | 71.82 | 4.13 | |
| S-GRPOBase Model=Qwen3-4B-Thinking-25072026.06 | 18,059 | 71.45 | 3.96 |