ResearchBenchmarksMulti-turn conversation on MT-Bench (Accuracy)Follow82.7AccuracyC2PO64.60469.3027478.698Dec 29, 2025Evaluation ResultsMethodMethodLinksAccuracyC2POBackbone=DeepSeek, Pro...Backbone=DeepSeek, Prompting Strategy=zero-shot2025.1282.7GPT-42025.1265.3