Graduate-level Science QA on GPQA D
5,318Tokens Used for ThinkingDRE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DREBase Model=Qwen3-30B-A3B-Thinking-25072026.06 | 5,318 | 75.51 | |
| DREBase Model=Qwen3-4B-Thinking-25072026.06 | 5,689 | 65.67 | |
| GRPOBase Model=Qwen3-30B-A3B-Thinking-25072026.06 | 6,236 | 74.68 | |
| GRPO + LPBase Model=Qwen3-30B-A3B-Thinking-25072026.06 | 6,435 | 75 | |
| DREBase Model=Qwen3-8B2026.06 | 6,658 | 60.98 | |
| Raw ModelBase Model=Qwen3-30B-A3B-Thinking-25072026.06 | 6,922 | 76.45 | |
| S-GRPOBase Model=Qwen3-30B-A3B-Thinking-25072026.06 | 7,000 | 74.68 | |
| GRPOBase Model=Qwen3-4B-Thinking-25072026.06 | 7,826 | 65.26 | |
| GRPO + LPBase Model=Qwen3-4B-Thinking-25072026.06 | 7,906 | 66.6 | |
| Raw ModelBase Model=Qwen3-4B-Thinking-25072026.06 | 8,091 | 65.47 | |
| GRPO + LPBase Model=Qwen3-8B2026.06 | 8,159 | 60.1 | |
| S-GRPOBase Model=Qwen3-4B-Thinking-25072026.06 | 8,435 | 65.59 | |
| GRPOBase Model=Qwen3-8B2026.06 | 8,474 | 60.61 | |
| Raw ModelBase Model=Qwen3-8B2026.06 | 8,683 | 59.34 | |
| S-GRPOBase Model=Qwen3-8B2026.06 | 9,660 | 59.79 |