ResearchBenchmarksAI-text explanation quality evaluation on AI-text detection 5000 samples (test)AI-text explanation quality evaluation on AI-text detection 5000 samples (test)Follow78.3Win RateTELL65.8269.0672.375.54May 27, 2026Evaluation resultsMethodMethodLinksWin Rate95% CITELLJudge=GPT-5.4-miniJudge=GPT-5.4-mini2026.0578.373.9TELLJudge=DeepSeek V4 FlashJudge=DeepSeek V4 Flash2026.0575.370.8TELLJudge=GPT-OSS 120BJudge=GPT-OSS 120B2026.0574.169.5TELLJudge=Panel meanJudge=Panel mean2026.0572.368.3TELLJudge=Gemma 4 26BJudge=Gemma 4 26B2026.0567.562.6TELLJudge=Nemotron SuperJudge=Nemotron Super2026.0566.361.5