Counterfactual error detection on Counterfactual dataset
24Finding Detection RateProReviewer
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ProReviewerBackbone=Qwen3-8B2026.06 | 24 | 29 | 27 | 27 | |
| AgentReviewBackbone=Qwen3.5-397B-A17B2026.06 | 10 | 16 | 20 | 15 | |
| AI Sci. V2Backbone=Qwen3.5-397B-A17B2026.06 | 9 | 26 | 27 | 21 | |
| DeepReviewBackbone=Qwen3-8B2026.06 | 3 | 2 | 9 | 5 | |
| CycleReviewerBackbone=Qwen3-8B2026.06 | 1 | 1 | 4 | 2 |