今日速览
| 序号 | 标题 | 来源 | 日期 | 主题 | 推荐等级 |
|---|---|---|---|---|---|
| 1 | Don’t Drop the BATON: Long-Horizon Robot Manipulation via Agentic Subtask Exploration and Transition-aware Memory | arXiv | 2026-08-17 | AI-Agent | 高 |
| 2 | When State Becomes an Attack Surface: State-Semantic Injection in LLM-Driven Embodied Agents | arXiv | 2026-08-17 | AI-Agent | 高 |
重点论文与技术动态
1. Don’t Drop the BATON: Long-Horizon Robot Manipulation via Agentic Subtask Exploration and Transition-aware Memory
- 来源：arXiv
- 日期：2026-08-17
- 作者/机构：Bingxin Xu, Yuzhang Shang, Emilio Ferrara
- 主题标签：
AI-Agent,arXiv - 推荐等级：高
- 分类：cs.RO, cs.AI, cs.CV
一句话结论
BATON通过子任务探索和过渡感知记忆显著提升长期机器人操作成功率。
一段话：BATON针对长期机器人操作中错误累积和子任务约束问题，提出将子任务作为探索单位，使探索成本从乘法（T^K）降至加法（T*K），失败可精准归因单个阶段；同时引入过渡感知记忆，通过验证代理控制VLA调用和跨子任务交接机制，确保子任务成功形式可被后续使用。在RoboMemArena基准测试中，任务成功率提升11.6%，累积成功率提升14.9%。
核心内容
- 长期任务中错误累积和子任务约束导致失败，现有方案依赖全任务探索，成本高且失败归因困难。
- BATON采用子任务探索机制，成本从乘法变为加法，失败可归因单个阶段。
- 过渡感知记忆包含验证代理控制调用过渡、跨子任务交接及前瞻过渡，确保子任务衔接顺畅。
方法与数据
- 方法：子任务探索机制与过渡感知记忆，无参数更新。
- 摘要未明确。
价值判断
- 值得关注：有效解决长期任务中的错误累积和过渡表示缺失问题。
- 可复用点：子任务探索框架可推广至其他多阶段机器人操作任务。
- 局限/待核查：摘要未明确。
摘要
Long-horizon robot manipulation chains many contact-rich skills into one multi-stage task. Vision-language-action (VLA) models increasingly master the individual skills, yet the chain still fails: errors compound beyond the policy's ability to correct, and one subtask silently constrains the next. A promising recipe freezes the VLA and puts an LLM agent in charge: it plans in language, moves in free space with analytic primitives, invokes the VLA only for contact-rich segments, and writes adaptation into language memory. Applied to long horizons, it breaks twice. (1) Competence comes from whole-task exploration at test time, whose cost is multiplicative in stages: if one stage needs T episodes, a K-stage task needs about T^K, and a failure does not reveal which stage caused it. (2) It has no representation of transitions: the VLA primitive carries an exit but no entry condition, so a subtask can succeed in a form its successor cannot use. We present BATON. Against (1), BATON makes the subtask the unit of exploration: each is explored in the cheap short-horizon regime and its solution stored in memory; a long-horizon trajectory is then composed from these solutions rather than discovered whole. Cost becomes additive (T*K) and every failure is attributed to a single stage. Against (2), BATON equips exploration with a transition-aware memory. Within a subtask, a verifier agent governs the invocation transition: the VLA is called only after the wrist view confirms the scene is ready. Across subtasks, a handoff transition restores an entry state disturbed by the predecessor's residue, and a lookahead transition selects the strategy whose outcome the successor can inherit. No parameters are updated. On the long-horizon benchmark RoboMemArena, BATON improves task success by 11.6% and cumulative success by 14.9% over the SoTA.2. When State Becomes an Attack Surface: State-Semantic Injection in LLM-Driven Embodied Agents
- 来源：arXiv
- 日期：2026-08-17
- 作者/机构：Jiawei Liu, Jiacheng Guo, Tian Zhang, Yiwei Xu, Juan Wang, Jinlin Fan
- 主题标签：
AI-Agent,arXiv - 推荐等级：高
- 分类：cs.RO, cs.AI
一句话结论
LLM驱动的具身智能体需整合场景状态信息以完成任务接地。
一段话：在LLM驱动的具身智能体中，模型不仅需理解用户指令，还需结合场景状态、对象属性、空间关系及执行反馈实现任务接地，随后将生成的行动规划交由技能库、运动规划器或控制器执行。
核心内容
- LLMs已从文本生成模型演变为具身智能体核心，支持环境感知、工具调用和任务执行，驱动其从纯文本处理向多模态交互发展。
- 现有系统如SayCan（任务推理）、Code as Policies（策略代码生成）、ProgPrompt（程序化提示）及VoxPoser（三维价值图构建）均利用LLMs进行任务规划与机器人操作。
- 智能体需融合多源信息（场景状态、对象属性等）完成任务接地，确保指令与环境动态匹配，避免执行偏差。
方法与数据
- 摘要未明确。
价值判断
- 值得关注：LLM与机器人系统的深度集成趋势，为具身智能体提供高效任务分解与规划能力。
- 可复用点：任务分解和推理框架（如SayCan、ProgPrompt）可直接应用于其他LLM驱动的具身智能体系统。
- 局限/待核查：摘要未提及具体局限或攻击细节（如标题暗示的“State-Semantic Injection”），需进一步验证。