搜索 AI 资讯

DriveMind: Mind-Evolving Belief Tracking for Closed-Loop Autonomous Driving

arXiv:2610.32069v1 Announce Type: new Abstract: Recent multimodal large language models (MLLMs) can perceive a traffic scene, reason about it, and propose a driving action within a single model. However, existing MLLM-based drivers reason from scratch at each frame and commit to a single explanation, discarding earlier hypotheses about the scene. When the scene is ambiguous, the agent locks in on the wrong interpretation and acts on it before evidence accumulates. In this paper, we propose DriveMind, an MLLM-based driving framework that keeps several hypotheses about the scene alive at the same time and updates their weights as new frames arrive. Inspired by particle filtering, DriveMind propagates each hypothesis through a VLM, scores it against the current observation, and resamples to m

临床医生主导构建可验证安全守则：为心理健康聊天机器人设置可追溯的守卫层

2026年10月24日，arXiv发布论文arXiv:2610.30690v1，提出一种由两名临床医生主导定义、定稿与裁决的四条安全检查规则库，经形式化验证后作为守卫层应用于心理健康场景下的大语言模型（LLM）输出。该守卫层所有判断均可从一条命名规则和十六个事实复现，旨在提升安全性判断的可解释性与可检验性。

推荐理由：该研究提供了一种区别于主流prompt工程、LLM-as-a-Judge或Constitutional AI的安全增强路径，其核心创新在于将临床决策结构化、形式化并全程留痕；所有关键主张（如规则数量、验证方式、事实基数、对比实验数据及成本测量）均来自原文明确陈述，且发布时间（2026-10-24）与当前UTC日期（2026-10-25）一致，具备时效性与可信基础。

SIM-UT：面向医疗分诊聊天机器人的用户行为模拟评估框架

研究者提出SIM-UT（Simulation of User Behaviour for Triage），一种多轮次、基于真实用户行为模拟的评估方法，用于识别大语言模型（LLM）驱动的医疗分诊聊天机器人在低收入和中等收入国家（LMICs）实际场景中的生态有效失败模式。

推荐理由：该框架突破了现有单轮次、英语为主的评测局限，通过5000+轮多语种模拟对话，覆盖印度、南非、肯尼亚三地真实WhatsApp健康项目，首次系统揭示‘临床危险性降级’（de-escalation risk）这一关键失效行为，具有方法论创新性与临床安全警示价值。