Hugging Face’s JAM model cuts robotics training costs by 73% without video generation
Hugging Face released a paper titled Vision-Language-Action or World-Action? Just Action Models, proposing a new approach called JAM that eliminates video generation during robotics training. The method uses a pretrained visual backbone and an action expert, trained via a foresight loss that predicts frozen future frame encodings—without retaining the generation branch at inference. 65.27%). The…
Key points
- JAM cuts robotics training FLOPs by 73% vs. FastWAM by caching foresight targets on 8 GH200 GPUs
- On LIBERO benchmark, JAM improves success rates from 97.0%/55.3%/91.2% to 98.2%/64.5%/91.4%
- Paper claims vision-language backbones and generative models perform equally well under JAM’s foresight loss
The paper argues the debate over vision-language-action (VLA) vs. world-action models (WAM) is overstated, as the performance gap isn’t tied to a single component—backbone initialization, future-observation objectives, or generation machinery. Cross-backbone tests on LIBERO-Plus show comparable results across Qwen3.5-2B (81.14%), Ovis-U1 (79.95%), and DINOv2-L (70.39%), all under JAM’s foresight recipe. The authors release public code built on FastWAM and ImageWAM, positioning JAM as a compute-efficient alternative to generative approaches in robotics.
The story so far
4 episodes →- Hugging Face’s JAM model cuts robotics training costs by 73% without video generationthis story
Hugging Face Paper: Robot Policies Don't Need Video Generation
aiweekly.co · 27 October 2026Loading the full article…
This text was published by aiweekly.co. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Robotics & Physical AI
All →- Trump targets foreign robot imports as national security threat · 2 src
- Robotics handbook highlights AI-driven humanoid and construction automation trends · 1 src
- Epic confirms AI writes Unreal Engine 5.9 code, but release uncertain · 2 src
- Spawn uses Fable-powered agents to autonomously build game world · 1 src
- OpenAI aims to use robots to build data centers by 2027 · 1 src
Comments
via GitHub Discussions