PapersMulti-sourceSelected10:47Chinese contentSTOCKTAKE:用公平预言衡量LLM智能体感知与行动差距这篇论文用STOCKTAKE基准把LLM智能体的‘没看懂’和‘看懂了却没行动’分开了,四个主流模型的表现差距比想象中大。#STOCKTAKE#Claude Sonnet 5#GPT-5.4#智能体2 个信源在谈EventaarXiv: DeepSeek@Sagar Deb, Ashwanth Krishnan3 sourcesSourceRead laterReadWorth followingUsefulFollow STOCKTAKE