ResearchBenchmarksUser-Agent Dialogue Evaluation on TRACE Easy (test)Follow93AccuracySCOPE85.7287.6189.591.39Oct 22, 2025Evaluation ResultsMethodMethodLinksAccuracyF1 ScorePrecisionRecallSCOPELLM/RM=Sonnet-3.5LLM/RM=Sonnet-3.52025.1093938898G-EvalLLM/RM=Sonnet-3.5LLM/RM=Sonnet-3.52025.10908981100SPURLLM/RM=Sonnet-3.5LLM/RM=Sonnet-3.52025.10909082100SCOPELLM/RM=GPT-4.1LLM/RM=GPT-4.12025.1090908399SPURLLM/RM=GPT-4.1LLM/RM=GPT-4.12025.10878778100RMLLM/RM=SkyworkLLM/RM=Skywork2025.1086848682G-EvalLLM/RM=GPT-4.1LLM/RM=GPT-4.12025.10868776100