SLIM-0.5B: Learning Action-Grounded Predictive Latents for Robot Manipulation
2026-08-10 • Robotics
Robotics
AI summaryⓘ
The authors present SLIM, a compact model that helps robots understand and act in their environment using fewer resources than larger models. Unlike big vision-language-action systems that focus on broad understanding, SLIM focuses on learning simple, efficient representations of actions and their effects. It does this by predicting future states and reconstructing actions from what it sees, all in a smaller, faster framework. Their tests show SLIM performs as well or better than bigger models while needing less memory and computation.
latent representationmasked trajectory predictionMixture-of-Transformersflow matchingvision-language-action policiesself-supervised learningrobot manipulationaction-conditioned predictionmultimodal modelspolicy learning
Authors
Jingkai Wang, Zihan Tang, Gu Zhang, Mingyu Cao, Jiapeng Chen, Jingjiao Zhao, Xiansheng Chen, Pengwei Wang, Lemao Liu, Dejing Dou
Abstract
Vision-language-action policies rely on large multimodal backbones to jointly perform perception, language conditioning, and action generation at every control step. Much of this capacity supports open-domain semantics, whereas continuous robot manipulation primarily requires compact representations of observations, actions, and the transitions induced by actions. Pixel-level world models provide another route, but predicting visual details irrelevant to control can be unnecessarily expensive. We propose SLIM (Self-supervised Latent Interaction Model), a compact 0.5B-parameter latent interaction policy. SLIM learns action-grounded predictive latents that capture both action-conditioned future transitions and the actions that explain observed changes. SLIM learns these representations through self-supervised masked trajectory prediction, combining action reconstruction with future-latent prediction. A compact Mixture-of-Transformers (MoT) backbone models interactions between observation latents and action tokens. The resulting policy is trained with flow matching for language-conditioned action generation. Across simulation benchmarks and real-world evaluation, SLIM matches or exceeds representative large-scale VLA and world-action-model baselines with fewer parameters, no additional embodied pretraining, lower inference latency, and substantially lower GPU memory usage.