Step-wise Verification on ProcessBench Overall
72.3F1 ScoreVerifySteer-Bi
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VerifySteer-BiBackbone=Qwen3-8B2026.05 | 72.3 | 21.23 | |
| Cons@4Backbone=Qwen3-8B2026.05 | 72.2 | 84.09 | |
| VerifySteer-UniBackbone=Qwen3-8B2026.05 | 71.7 | 21.23 | |
| Qwen3-14BBackbone=Qwen3-14B2026.05 | 71.6 | 37.45 | |
| Prompt Opt.Backbone=Qwen3-8B2026.05 | 69.9 | 21.26 | |
| Basic PromptBackbone=Qwen3-8B2026.05 | 65.9 | 16.71 | |
| Qwen2.5-72BBackbone=Qwen2.5-72B2026.05 | 59.7 | 53.03 | |
| Llama-3.3-70BBackbone=Llama-3.3-70B2026.05 | 58.6 | 85.45 | |
| VerifySteer-BiBackbone=Qwen3-1.7B2026.05 | 29.7 | 3.47 | |
| VerifySteer-UniBackbone=Qwen3-1.7B2026.05 | 29 | 3.37 | |
| Cons@4Backbone=Qwen3-1.7B2026.05 | 26 | 12.71 | |
| Prompt Opt.Backbone=Qwen3-1.7B2026.05 | 25.1 | 3.19 | |
| Basic PromptBackbone=Qwen3-1.7B2026.05 | 20.9 | 2.86 |