PapersMulti-source11:13Chinese contentFisher-R1:训练LLM代理实现可靠假设检验开源模型Fisher-R1专治假设检验,比GPT-5.4和DeepSeek-V4-Pro都强,最难任务能提26%,搞数据分析的可以看看。#Fisher-R1#P-Bench#hypothesis testing#agents4 sourcesEventaarXiv cs.AI@Jiacheng Miao et al. (4)SourceRead laterReadWorth followingUsefulFollow Fisher-R1