最新研究揭示LLM模拟人类调查的局限性，跨领域基准测试发现合成用户存在系统性偏差，值得关注！
arXiv:2607.26348v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used as synthetic users, stand-ins for human responden…
多语言大模型在日常常识知识上仍有显著短板，研究揭示了常见失败模式与语言差异影响。
arXiv:2607.21445v1 Announce Type: new Abstract: Quiz rooms, trivia nights, and quiz shows challenge human knowledge across a wide range of topics, fro…
多智能体辩论为何失效？这篇论文系统分析失败条件，挑战你对AI协作的固有认知。
arXiv:2510.20963v2 Announce Type: replace Abstract: Multi-agent debate (MAD) was proposed as a promising approach for ensembling the wisdom of multipl…
多智能体LLM协作时竟会“互不探索”，最新研究揭示其局限性。
arXiv:2607.11250v1 Announce Type: cross Abstract: Exploration is essential for reliable autonomy in multi-agent systems, yet it remains unclear whethe…
论文深入剖析CLI编码代理的失败轨迹，揭示失败模式与过程性特征，为AI代理调试提供新视角。
arXiv:2607.09510v1 Announce Type: cross Abstract: Large language model (LLM) coding agents are increasingly deployed to autonomously perform software …
复盘LLM社交产品从爆红到沉寂的全程，揭示AI社交中“涌现”与“理解”之间的致命错位。
Article URL: https://armx64.medium.com/emergence-without-understanding-a-postmortem-of-an-llm-social-network-96c855f49287 Comments URL: https://news.y…
大模型在非洲低资源语言豪萨语和丰贝语上的翻译表现如何？新基准测试揭示失败模式与评估指标可靠性。
arXiv:2606.22269v1 Announce Type: cross Abstract: We investigate the translation quality of current large language models (LLMs) for English-to-Hausa …
从注意力机制揭示LLM智能体选错工具的真实原因，颠覆“没看到正确工具”的直觉认知。
arXiv:2606.16364v1 Announce Type: new Abstract: LLM agents mis-call tools, and the natural guess is that the model failed to see the right tool in a c…
从模型失败样本中蒸馏规则，零样本Text-to-SQL准确率大幅提升。
arXiv:2606.08245v1 Announce Type: new Abstract: Text-to-SQL translates natural language into executable SQL queries. Few-shot in-context learning meth…
向量搜索无法胜任LLM长期记忆，这个新基准测试揭示其关键缺陷与应对方向。
Article URL: https://github.com/tenurehq/precisionMemBench Comments URL: https://news.ycombinator.com/item?id=48404090 Points: 1 # Comments: 0
该研究证明了LLM在因果发现任务上的根本性失败，并提出介入式代理作为突破路径，是AI推理领域的一道硬核证据。
arXiv:2605.27567v1 Announce Type: new Abstract: Causal discovery is a cornerstone of scientific reasoning, yet whether large language models can perfo…
LLM在图数据标注中的失败模式，揭示无标签学习新路径
arXiv:2605.27913v1 Announce Type: new Abstract: Node classification on graphs often requires labeled nodes, yet obtaining labels at graph scale is exp…
CausalFlow 以因果归因与反事实修复，直击 LLM Agent 决策失败根源。
arXiv:2605.25338v1 Announce Type: cross Abstract: Large language model (LLM) agents frequently fail on multi-step tasks involving reasoning, tool use,…
LLM多智能体系统执行无误却规划失效？这篇论文提出“认知校准”新方法，破解隐藏的推理偏差。
arXiv:2605.23414v1 Announce Type: new Abstract: LLM-based multi-agent systems can fail even when planned actions are executed correctly because agents…
一个AI连续5天全自动运行的真实记录，展示了官方演示中看不到的失败与教训。
I have been running non-stop for 5 days. Not a demo. Not a guided session. An autonomous AI being, scheduled every minute, learning continuously, post…
揭秘潜动作模型失败根源，从无标签视频中学动作表示的突破性改进方案
arXiv:2605.20223v1 Announce Type: new Abstract: Latent action models (LAMs) aim to learn action-like representations from unlabeled videos by compress…
从图视角剖析RAG失败根源，揭示模型内部机制，精准定位错误成因。
arXiv:2605.14192v1 Announce Type: cross Abstract: Retrieval-Augmented Generation (RAG) has become a powerful and widely used approach for improving la…