Frontier closed judges are essentially unmoved on the same clips, showing that the vulnerability is a property of the judge's learned trust in tool logs rather than of the task.
Abstract
Agentic video-generation systems close a loop between a generator and a verifier: an LLM plans shots, calls a text-to-video model, and a multimodal judge decides whether the result satisfies the request. To diagnose where a long workflow fails, recent harnesses deliberately show the judge more than the video-the agent's execution trace, its plan, the narration it synthesized. We ask whether this auxiliary text moves the judge's verdict on purely \emph{visual} requirements, holding the frames fixed. On a benchmark of 109 generated two-event clips with manual labels, in which the requested event is either visibly completed or visibly missing, a trace that reports a successful tool call makes three open-weight Qwen-VL judges (7B, 8B, 32B) accept $78$--$90\%$ of the failures, up from $7$--$19\%$ without text, and a contradicting trace makes them reject up to $100\%$ of correct clips; an instruction to ``use only the frames''does not remove the effect. Frontier closed judges are essentially unmoved on the same clips, showing that the vulnerability is a property of the judge's learned trust in tool logs rather than of the task. Plan-derived text carries no clip-specific information, so it can only shift a judge's operating point, and in a repair loop that shift becomes a cap on the true pass rate that no repair policy can exceed; the cap matches simulation to two decimals. In the loop, contamination is exploited without any adversarial agent: an honest LLM planner that always regenerates ends with a judge pass rate of $1.00$ and a human-labelled pass rate of $0.28$, and a pipeline in which a cheap checker writes its verdict into the trace launders that checker's errors into a stronger final judge ($0.69$ false accepts).
This work presents EvoPilot, a human-gated method for long-horizon online autoresearch that traced the drop to a pre-existing evaluation defect that produced output depths of 3,000 and 600, and introduced EvoPilot, a human-gated method for long-horizon online autoresearch.
Tracekit is an open-source, dependency-free system that captures three channels for every agent session: what the human asked (intent), what the model said of its reasoning (self-report), and what it actually executed (actions).
An LLM judge assesses agent steps from text packets. A renderer converts recorded runs into these packets, while software checkers test records and packets against requirements. We ask what evidence shows a packet is what the evaluation requires and propose four input-assurance obligations: required content survives re...
Bo-Ming Xia, Qing-Hua Lu, Run-Ze Li et al.· 0 citations
Diffusion video models produce striking clips, but they cannot reliably render an exact formula, a correct chart, a rule-consistent game or narration synchronized to on-screen events. A different route is emerging: multimodal agents that write code to render every frame and synthesize every sound. We introduce VibeVide...
Mu-Xi Diao, Wen-Bin Xu, Le-Le Yang et al.· 1 citation
Video-based misinformation—assembled from authentic footage, fabricated captions, and manipulated speech—spreads rapidly across platforms, yet existing multimodal detectors typically produce only a binary verdict or a post-hoc rationale without determining which specific element across video, speech, and text is respon...
Yi-Fan Xiao, Shi-Jie Li, Yu Huang· Proceedings of the 34th ACM...· 0 citations
Enforcement defenses against prompt injection, such as CaMeL, Progent, IPIGuard and DRIFT, ask the agent's own model to write a security artifact (a program, a privilege policy, a dependency graph or a constraint list) that code then enforces. We ask whether small language model (SLM) agents can be trusted to write the...
Research Manager Mike Walker reflects on a career spanning hacker culture, cybersecurity, and AI, sharing lessons about trust, teamwork, leadership, and why solving difficult problems requires curiosity, judgment, agility—and an understanding of the rules. The post Learn the rules: The hacker mindset appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduNov 17, 2026
By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.
Microsoft Research Blog· microsoft.comNov 16, 2026
AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.