English(EN) When Can You Poison Rewards? A Tight Characterization of Reward Poisoning in Linear MDPs

新研究精确刻画了强化学习中的奖励投毒漏洞

作者 PulseAugur 编辑部 · [1 个来源] · 2026-07-07 04:00

本文对强化学习智能体何时易受奖励投毒攻击进行了精确刻画。研究聚焦于线性马尔可夫决策过程（MDP），并为对手在有限预算内成功操纵智能体策略建立了必要和充分条件。该框架通过将环境近似为线性MDP，可以扩展到深度强化学习，证明了其在识别和利用可攻击智能体方面的理论和实践意义。 AI

影响为理解和缓解强化学习系统中的对抗性攻击提供了理论基础。

排序理由该集群包含一篇详细介绍强化学习漏洞理论研究的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍强化学习漏洞理论研究的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, safety; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: 93 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

arXiv cs.LG TIER_1 English(EN) · Jose Efraim Aguilar Escamilla, Haoyang Hong, Jiawei Li, Haoyu Zhao, Xuezhou Zhang, Sanghyun Hong, Huazheng Wang · 2026-07-07 04:00

何时可以进行奖励投毒？线性 MDP 中奖励投毒的精确刻画

arXiv:2604.10062v3 Announce Type: replace Abstract: We study reward poisoning attacks in reinforcement learning (RL), where an adversary manipulates rewards within constrained budgets to force the target RL agent to adopt a policy that aligns with the attacker's objectives. Prior…