Logical Reasoning on Countdown CD4
59.4Avg@16GPS
Evaluation Results
| Method | Links | |
|---|---|---|
| GPSBackbone=Qwen3-8B, Finetuning=Yes, Runtime=5.0h2026.02 | 59.4 | |
| DSBackbone=Qwen3-8B, Finetuning=Yes, Runtime=6.9h2026.02 | 58.7 | |
| GPSBackbone=Qwen3-4B, Finetuning=Yes, Runtime=3.4h2026.02 | 57.2 | |
| DSBackbone=Qwen3-4B, Finetuning=Yes, Runtime=4.9h2026.02 | 56.1 | |
| MoPPSBackbone=Qwen3-8B, Finetuning=Yes, Runtime=4.3h2026.02 | 55.5 | |
| GRESOBackbone=Qwen3-8B, Finetuning=Yes, Runtime=6.2h2026.02 | 54.1 | |
| GRESOBackbone=Qwen3-4B, Finetuning=Yes, Runtime=4.9h2026.02 | 53.8 | |
| PCLBackbone=Qwen3-8B, Finetuning=Yes, Runtime=5.1h2026.02 | 53.5 | |
| MoPPSBackbone=Qwen3-4B, Finetuning=Yes, Runtime=2.8h2026.02 | 52.9 | |
| UniformBackbone=Qwen3-8B, Finetuning=Yes, Runtime=4.3h2026.02 | 52.5 | |
| UniformBackbone=Qwen3-4B, Finetuning=Yes, Runtime=2.9h2026.02 | 51.1 | |
| PCLBackbone=Qwen3-4B, Finetuning=Yes, Runtime=3.2h2026.02 | 51 | |
| Qwen3-8BBackbone=Qwen3-8B, Finetuning=None2026.02 | 2.1 | |
| Qwen3-4BBackbone=Qwen3-4B, Finetuning=None2026.02 | 1.3 |