系统化诊断长时程安全LLM代理的失败根因，突破端到端指标的盲区。
arXiv:2608.20563v1 Announce Type: cross Abstract: Long-horizon security LLM agents must carry information and decisions across many dependent interact…
诊断Video-LLM在长视频中跟踪角色的失败，揭示当前最强开源模型未能真正“观看”视频的局限。
arXiv:2607.11078v1 Announce Type: cross Abstract: Can a Video Large Language Model (Video-LLM) follow one person through a long video, keeping track o…
新基准ToolFailBench精准诊断LLM agent工具调用失败原因，揭示聚合分数下的隐藏缺陷。
arXiv:2607.04686v1 Announce Type: cross Abstract: Tool calling is central to modern language model agents, but aggregate benchmark scores often hide w…
被ICML 2026收录，提出逐步置信度归因方法，精准诊断黑盒大模型的多步推理失败原因。
arXiv:2605.19228v1 Announce Type: cross Abstract: Large Language Models have achieved strong performance on reasoning tasks with objective answers by …