今日速览
| 序号 | 标题 | 来源 | 日期 | 主题 | 推荐等级 |
|---|---|---|---|---|---|
| 1 | Recognition Without Enforcement: Configuration-Dependent Failures in LLM Agent Instruction Arbitration and External Control | arXiv | 2026-08-28 | RAG | 高 |
| 2 | LUCID: An Agentic AI Framework on Digital-Twin in the Loop for QoS-Guaranteeing Robotic Control | arXiv | 2026-08-28 | AI-Agent | 高 |
重点论文与技术动态
1. Recognition Without Enforcement: Configuration-Dependent Failures in LLM Agent Instruction Arbitration and External Control
- 来源:arXiv
- 日期:2026-08-28
- 作者/机构:Jun Wen Leong
- 主题标签:
RAG,arXiv - 推荐等级:高
- 分类:cs.CR
一句话结论
安全代理需要外部强制,而非仅依赖更好的识别。
一段话:LLM代理在指令仲裁中存在识别-执行差距,模型能解码源格式特征(如角色模板位置、通道元数据)并识别伪造权威,但特定配置仍导致冲突工具调用;漏洞集中在可重现单元,随部署窗口动态变化(窗口内最高47个百分点);外部参考监控通过认证源路由与能力门控工具执行,确定性拒绝所有测试的伪造、篡改、重放及未签名请求。
核心内容
- 识别-执行差距:源格式特征可线性解码,模型能识别伪造权威,但配置错误时仍执行冲突工具调用。
- 漏洞动态集中:平均执行率仅1.21%(14,294次试验),漏洞集中在可重现单元,部署窗口内变化显著。
- 外部监控有效:认证源路由结合能力门控工具执行,确定性拒绝所有测试的伪造请求。
方法与数据
- 舰队评估覆盖46个模型端点(6供应商,含开源权重)和48个模型,平均执行率1.21%。
价值判断
- 值得关注:安全代理需外部强制而非仅改进识别能力。
- 可复用点:外部参考监控架构(认证源路由+能力门控)可复用于其他安全场景。
- 局限/待核查:提示层防御不泛化;曾发现实现缺陷(时钟偏移接受,已修补)。
摘要
LLM agents arbitrate among instructions from system prompts, users, memory, and tools, but this arbitration cannot be assumed to enforce trust boundaries. We identify a recognition-enforcement gap: source-format features (role-template position, channel metadata, formatting cues) are linearly decodable from model activations, and models can explicitly identify forged authority when prompted, yet some configurations still produce the conflicting tool call. We use "recognition" in this specific decodable-source-format-plus-verbalized-detection sense; crossed-probe controls show it is not a unified abstract trust representation. The gap is not an immutable property of model weights. Restrictive policies and diverse prompts can eliminate execution on the same models, while permissive configurations and particular prompt-model pairs yield deterministic failures. Across a fleet evaluation (authority spoofing: 46 model endpoints across 6 vendors including open-weight; memory conflict: 48 models), average execution under diverse novel attacks is 1.21% [0.5-2.1%] (model-clustered CI over 14,294 spoofed trials from 29 models), but vulnerability is concentrated in reproducible cells and shifts across deployment windows (up to 47pp within-window per-fingerprint range). Prompt-layer defenses likewise fail to generalize across models and adaptive formulations. We therefore treat model self-arbitration as a capability rather than a security boundary and implement an external reference monitor combining authenticated source routing with capability-gated tool execution. It deterministically rejects all tested forged, tampered, replayed, and unsigned requests while preserving legitimate operations. A separate adaptive red-team found one implementation flaw (a since-patched clock-skew admission), not a cryptographic bypass. Secure agents require external enforcement, not merely better recognition.2. LUCID: An Agentic AI Framework on Digital-Twin in the Loop for QoS-Guaranteeing Robotic Control
- 来源:arXiv
- 日期:2026-08-28
- 作者/机构:Hyeonsu Lyu, Minwoo Kim, Sehyun Ryu, Hyun Jong Yang
- 主题标签:
AI-Agent,arXiv - 推荐等级:高
- 分类:eess.SY, cs.RO
一句话结论
LUCID通过动态协调轨迹规划与无线资源管理,实现云机器人系统在动态环境中的QoS保障。
一段话:LUCID将轨迹规划和无线资源管理从固定优化问题转变为动态配置的优化问题模式,由操作员高级意图驱动;利用SimBridge将大规模机器人场景转换为无线就绪数字孪生,实现高效无线冲突评估;并通过FastConfigNet降低规划延迟,实验验证其在变化意图、机器人数量和场景中的稳健适应性。
核心内容
- 现有方法将TP-RRM视为固定优化问题,无法动态适应环境变化,导致QoS违规。
- LUCID将TP-RRM公式化为有界模板,动态配置变量、目标和约束以响应操作员意图。
- SimBridge支持重复光线追踪评估,FastConfigNet显著减少规划延迟。
方法与数据
- LLM代理协调的数字孪生在环(DITL)环境用于动态优化问题模式。
- 摘要未明确。
价值判断
- 值得关注:解决云机器人动态环境下的QoS保障核心挑战。
- 可复用点:动态优化问题模式协调机制可推广至其他实时系统。
- 局限/待核查:摘要未明确。