Error Detection on Perturbation Benchmark 24-paper frontier
0.716RecallOpenAIReview
Evaluation Results
| Method | Links | |
|---|---|---|
| OpenAIReviewBackbone=GPT-5.52026.06 | 0.716 | |
| OpenAIReviewBackbone=Claude-Opus-4.72026.06 | 0.681 | |
| GPT-5.5System=zero-shot2026.06 | 0.598 | |
| OpenAIReviewBackbone=DeepSeek-V4-Flash2026.06 | 0.558 | |
| Claude-Opus-4.7System=zero-shot2026.06 | 0.543 | |
| OpenAIReviewBackbone=Grok-4.1-Fast2026.06 | 0.526 | |
| OpenAIReviewBackbone=Qwen3.6-35B-A3B2026.06 | 0.454 | |
| OpenAIReviewBackbone=Gemini-3.1-Flash-Lite2026.06 | 0.33 | |
| Grok-4.1-FastSystem=zero-shot2026.06 | 0.314 | |
| Qwen3.6-35B-A3BSystem=zero-shot2026.06 | 0.314 | |
| DeepSeek-V4-FlashSystem=zero-shot2026.06 | 0.31 | |
| Reviewer3 (closed)2026.06 | 0.265 | |
| DeepSeek-V4-FlashSystem='coarse2026.06 | 0.207 | |
| Qwen3.6-35B-A3BSystem='coarse2026.06 | 0.182 | |
| Gemini-3.1-Flash-LiteSystem=zero-shot2026.06 | 0.147 | |
| Grok-4.1-FastSystem='coarse2026.06 | 0.128 | |
| Gemini-3.1-Flash-LiteSystem='coarse2026.06 | 0.123 |