PapersBuilders11:29Chinese content研究比较六种免执行代码正确性预测器在分布偏移下的表现让模型自己判断代码对不对,又不想跑测试?这篇论文测了六种免执行预测器在换语言、换基准后还靠不靠谱,结论很具体。#Qwen3-Coder#gpt-oss#code correctness#AUROCaarXiv cs.LG@Viola Campos et al. (3)SourceRead laterReadWorth followingUsefulFollow Qwen3-Coder
PapersBuildersOfficial09:01Chinese contentAnthropic:LLM 让形式化软件验证得以大规模落地Anthropic 讲了怎么用 LLM 搞形式化验证,写代码的朋友可以看看代码正确性能不能这么保证。#Anthropic#formal verification#LLM#code correctness7 sourcesEventAPrimary sourceAnthropic: ResearchSourceRead laterReadWorth followingUsefulFollow Anthropic
Papers11:11Chinese contentQwen3代码正确性在隐藏状态中的可解码性论文揭示Qwen3隐藏状态如何预测代码正确性#Qwen3#code correctness#hidden states#LiveCodeBenchaarXiv cs.LG@Carlo Di CiccoSourceRead laterReadWorth followingUsefulFollow Qwen3