1
TsuGO: Probing Search Efficiency in LLM Reasoning via Go Life-and-Death Problems
用围棋死活题当试金石,揭开大模型搜索推理效率的底牌,视角新鲜且硬核。
arXiv:2608.13221v1 Announce Type: new Abstract: The evaluation of LLM reasoning is moving from final-answer accuracy to process-level assessment, yet …