Execution Before Memory: Diagnosing Memory Failures in Video World Models
A memory test for a video world model should say whether the scene was set, whether the asked change was carried out, and whether the last frame shows the result.
ProofPaper ↗
Key points
- When the change is not carried out, the hidden state transition cannot appear, so the test is not really a test of memory.
- We introduce MIST, an evaluation that separates the change from the memory it should leave behind.
- The same gap appears in general memory evaluations.
- On that basis we propose M3-Agent (Make the Missing Move), which recursively revises the details of the input instruction so that the model successfully carries out the requested instruction.
Sources (2)
- [1]Execution Before Memory: Diagnosing Memory Failures in Video World ModelsHugging Face Daily Papers · Oct 16, 12:00 AM
A memory test for a video world model should say whether the scene was set, whether the asked change was carried out, and whether the last frame shows the result.
When the change is not carried out, the hidden state transition cannot appear, so the test is not really a test of memory.
- [2]Execution Before Memory: Diagnosing Memory Failures in Video World ModelsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 16, 02:37 PM · same content
Extractive summary: sentences quoted from the sources.
Before this
- Oct 15, 2026RoleQ: Role-Aligned Scales for Converting Language Models to Ternary
- Oct 15, 2026Anthropic tells investors quarterly revenue up nearly 50% to $16.8bn ahead of IPO
- Oct 15, 2026GeoState: Learning View-Decoupled Geometry States for 3D World Modeling
- Oct 15, 2026CoWorld Reasoner: Agentic Multimodal Reasoning with the Evolving Video World Model-Driven Harness
- Oct 15, 2026Velox: Spatial Memory Makes World Models Efficient
- Oct 14, 2026Sparse Transformer: a Controlled Testbed for Mechanistic Interpretability
