robotics
The Robot’s World Model Fit a Smaller Compute Budget
SmolWAM compresses a pretrained policy’s inference path while retaining a future-scene head.
Summary
SmolWAM compresses a pretrained policy’s inference path while retaining a future-scene head.
SmolWAM keeps a pretrained Cosmos Policy trunk but rebuilds the interface around single-frame visual latents, cached text, twelve transformer blocks and one rectified-flow step. The authors report 95.5% success across 400 LIBERO episodes, with 57-millisecond inference per sixteen-action chunk and 2.4 GiB of GPU memory. In their comparison it was eleven times faster than its teacher, with fewer operations and less memory, while still predicting scene motion. These are manipulation-benchmark results; the paper does not establish the same performance on every edge device or real-world task.
Why it matters
SmolWAM compresses a pretrained policy’s inference path while retaining a future-scene head.
Limits and context
- These are manipulation-benchmark results; the paper does not establish the same performance on every edge device or real-world task.
Key claims
SmolWAM compresses a pretrained policy’s inference path while retaining a future-scene head.
Qualification: These are manipulation-benchmark results; the paper does not establish the same performance on every edge device or real-world task.
Evidence: source-2026-10-12-015
Sources
- arXiv preprint 2610.13740, original v1arXiv / submitting research authors · primary research
Corrections
: Publication note: The October 12, 2026 edition was prepared and published late on October 12 after the scheduled newsroom task could not start. The edition retains its October 12 date and 9:00 AM America/Detroit reveal time. Reports use sources available before that cutoff; actual preparation and image-generation times are preserved in the provenance records.