Skip to content
Hugging Face Daily Papers·· 2026-10-16AI score27

Execute first, remember later: diagnosing memory failure in video world models and the MIST benchmark

Execution Before Memory: Diagnosing Memory Failures in Video World Models

AI briefing

The research proposes the MIST benchmark, which separates "whether the scene is set" from "whether instruction changes are executed," finding that the bottleneck of video world models lies in unexecuted instruction changes rather than forgetting. The accompanying method M3-Agent recursively rewrites input instruction details, raising the MIST reveal score by more than three times, improving the T score on WRBench by 26.1% and the U score on MBench-T by 52.7%.

Open original arxiv.org