List your company

Error Detection on Perturbation Benchmark 24-paper frontier

0.716RecallOpenAIReview
0.099280.259390.41950.57961Jun 18, 2026

Evaluation results

MethodLinks
OpenAIReviewBackbone=GPT-5.52026.060.716
OpenAIReviewBackbone=Claude-Opus-4.72026.060.681
GPT-5.5System=zero-shot2026.060.598
OpenAIReviewBackbone=DeepSeek-V4-Flash2026.060.558
Claude-Opus-4.7System=zero-shot2026.060.543
OpenAIReviewBackbone=Grok-4.1-Fast2026.060.526
OpenAIReviewBackbone=Qwen3.6-35B-A3B2026.060.454
OpenAIReviewBackbone=Gemini-3.1-Flash-Lite2026.060.33
Grok-4.1-FastSystem=zero-shot2026.060.314
Qwen3.6-35B-A3BSystem=zero-shot2026.060.314
DeepSeek-V4-FlashSystem=zero-shot2026.060.31
Reviewer3 (closed)2026.060.265
DeepSeek-V4-FlashSystem='coarse2026.060.207
Qwen3.6-35B-A3BSystem='coarse2026.060.182
Gemini-3.1-Flash-LiteSystem=zero-shot2026.060.147
Grok-4.1-FastSystem='coarse2026.060.128
Gemini-3.1-Flash-LiteSystem='coarse2026.060.123