ResearchBenchmarksSelf-doubt detection on Original Source DatasetsFollow84.06Self-Doubt AUROCSELFDOUBT68.09672.240576.38580.5295Apr 7, 2026Evaluation ResultsMethodMethodLinksSelf-Doubt AUROCSELFDOUBTModel=gpt-oss-120bModel=gpt-oss-120b2026.0484.06SELFDOUBTModel=Claude Sonnet 4.6Model=Claude Sonnet 4.62026.0483.58SELFDOUBTModel=Grok 4.1 FastModel=Grok 4.1 Fast2026.0481.01SELFDOUBTModel=gpt-oss-20bModel=gpt-oss-20b2026.0479.08SELFDOUBTModel=Qwen3-14BModel=Qwen3-14B2026.0478.76SELFDOUBTModel=Qwen3Model=Qwen32026.0477.45SELFDOUBTModel=Gemini 2.5 FlashModel=Gemini 2.5 Flash2026.0468.71