Reverse Chain-of-Thought Generation on IFEval
85.1AccuracySSR-D
Evaluation Results
| Method | Links | |
|---|---|---|
| SSR-DBackbone=Qwen3-32B-Think2026.02 | 85.1 | |
| SSRBackbone=Qwen3-32B-Think2026.02 | 84.8 | |
| SUPBackbone=Qwen3-32B-Think2026.02 | 84.5 | |
| SSR-DBackbone=Qwen3-4B-Think-25072026.02 | 83.8 | |
| SSR-DBackbone=Qwen3-8B-Think2026.02 | 83.7 | |
| NEUBackbone=Qwen3-32B-Think2026.02 | 83.4 | |
| AUG-SUPBackbone=Qwen3-32B-Think2026.02 | 83.2 | |
| SSRBackbone=Qwen3-8B-Think2026.02 | 82.9 | |
| SSRBackbone=Qwen3-4B-Think-25072026.02 | 81.6 | |
| AUG-SUPBackbone=Qwen3-8B-Think2026.02 | 81.5 | |
| SUPBackbone=Qwen3-8B-Think2026.02 | 81.2 | |
| AUG-SUPBackbone=Qwen3-4B-Think-25072026.02 | 80.1 | |
| NEUBackbone=Qwen3-8B-Think2026.02 | 80 | |
| SUPBackbone=Qwen3-4B-Think-25072026.02 | 78.5 | |
| SSR-DBackbone=NBG4-3B-Base2026.02 | 78.2 | |
| SSRBackbone=NBG4-3B-Base2026.02 | 77.6 | |
| AUG-SUPBackbone=NBG4-3B-Base2026.02 | 77.2 | |
| NEUBackbone=Qwen3-4B-Think-25072026.02 | 76.7 | |
| SUPBackbone=NBG4-3B-Base2026.02 | 76.5 | |
| NEUBackbone=NBG4-3B-Base2026.02 | 74.3 |