DRIFT: Derailing Denoising Trajectories of Flow-Matching VLAs with Adversarial Patch Attack

• 类型：arxiv
• 标识：2608.03207
• 链接：https://arxiv.org/abs/2608.03207
• 主分类：multimodal
• 形态：method
• 被引：0
• 被引来源：Semantic Scholar
• S2被引：0
• 影响力被引：0
• TLDR：This work introduces DRIFT (Denoising Redirection via Input perturbation of the Flow-matching Trajectory), a test-time universal adversarial patch placed on the robot's gripper that attacks the denoising velocity field of an off-the-shelf policy.
• 副分类：risk
• 待LLM分类：否
• 标题中文：DRIFT：利用对抗性 patch 攻击偏转流匹配 VLA 的去噪轨迹
• TLDR中文：提出 DRIFT（Denoising Redirection via Input perturbation of the Flow-matching Trajectory），一种测试期通用对抗补丁，贴在机器人夹爪上，针对现成策略的去噪速度场进行攻击。
• 来源文件：
• /inbox/tom/_candidates/2026-08-06-agent-rag-longcontext-candidates.json
• /inbox/tom/_candidates/2026-08-07-agent-rag-longcontext-candidates.json
• [S2 enrich]