Execution Before Memory: Diagnosing Memory Failures in Video World Models
Merged summary · from the lead articleSummaries by Vani AI
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. Current tests fold these into one number, so a low score is easy to read as forgetting. 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…
Coverage
oldest first · one row per article · UTC- · 27d ago
Hugging Face daily papersFirst report
MIST finds video world models fail to execute changes, not only memory
M3-Agent more than triples the MIST reveal score by revising instructions so a frozen model executes the change.
- · 27d ago
PPMem stores video world-model scenes as compact planar memory
About 100K plane primitives, roughly 5MB per scene, stay coherent across hundreds of frames under large viewpoint changes.