DeepSeek 3.1 Terminus

A new open-source benchmark framework, ParliamentBench, based on the social deduction game Secret Hitler, evaluates large language models' deceptive capabilities, finding that frontier models GPT-5.4,…

A new open-source benchmark framework, ParliamentBench, based on the social deduction game Secret Hitler, evaluates whether large language models (LLMs) can deceive, persuade, and reason under informa…