← 研踪 PaperTrailarXiv:2606.23892 · v2
RedPhys: An Extensible Red-Teaming Platform for Physical-World Models and Agents
首次提交 2026-06-23 03:41
版本更新 2026-11-08 01:20
雷达观测 2026-11-10 11:14
摘要
Physical-world models (PhyModels) perform grounded perception, reasoning, or decision tasks; physical-world agents (PhyAgents) combine these models with planning or tool use. We introduce RedPhys, an extensible red-teaming platform, instantiated here with vision-language models and agents. Shared interfaces make tasks, test methods, defenses, and victim systems configurable and extensible, producing robustness reports and datasets for future alignment. The platform emphasizes adversarial testing that seeks physically grounded functional failures, while also supporting non-adversarial input perturbations without an attack objective. The benchmark integrates 12 red-teaming methods, 3 model-agnostic defenses, and 13 PhyModels across seven physical domains under a practical black-box threat model. An agentic target-generation pipeline derives a scene-specific, physically grounded target, so targeted methods pursue the same intended error. Our evaluation shows that when attackers control the text prompt or add auxiliary visual instructions, injection exposes substantial instruction-following vulnerabilities; among bounded visual-transfer attacks, multimodal co-optimization transfers best; single-pass generative attacks approach iterative methods without per-image optimization; and neither model scale nor physical-reasoning post-training confers robustness. Code is available at https://github.com/loopleap-lab/RedPhys.
作者备注：Accepted at NeurIPS 2026 Evaluations and Datasets Track. Code: https://github.com/loopleap-lab/RedPhys
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v2 · 2026-11-08 01:20 · 最新观测版本
RedPhys: An Extensible Red-Teaming Platform for Physical-World Models and Agents
Physical-world models (PhyModels) perform grounded perception, reasoning, or decision tasks; physical-world agents (PhyAgents) combine these models with planning or tool use. We introduce RedPhys, an extensible red-teaming platform, instantiated here with vision-language models and agents. Shared interfaces make tasks, test methods, defenses, and victim systems configurable and extensible, producing robustness reports and datasets for future alignment. The platform emphasizes adversarial testing that seeks physically grounded functional failures, while also supporting non-adversarial input perturbations without an attack objective. The benchmark integrates 12 red-teaming methods, 3 model-agnostic defenses, and 13 PhyModels across seven physical domains under a practical black-box threat model. An agentic target-generation pipeline derives a scene-specific, physically grounded target, so targeted methods pursue the same intended error. Our evaluation shows that when attackers control the text prompt or add auxiliary visual instructions, injection exposes substantial instruction-following vulnerabilities; among bounded visual-transfer attacks, multimodal co-optimization transfers best; single-pass generative attacks approach iterative methods without per-image optimization; and neither model scale nor physical-reasoning post-training confers robustness. Code is available at https://github.com/loopleap-lab/RedPhys.
查看此版本解读 →查看此版本原文 ↗