Backend
Latent Action Policies from Suboptimal Data
arXiv stat.ML0 views
arXiv:2611.24482v1 Announce Type: cross Abstract: Latent Action Models (LAMs) learn control representations from action-free video, but existing pipelines typically train latent policies with behavioral cloning, without an explicit policy improvement objective. Although offline RL offers a natural approach to policy improvement, standard methods do not transfer directly to learned latent action spaces. We find that the structure of learned latent actions makes naive offline RL unreliable: directly maximizing value over latent actions or fitting a unimodal policy can produce latent actions that cannot be realized by physical execution, while the latent representation can merge physical controls that receive different rewards. To understand when latent-space optimization remains valid, we derive conditions under which policy improvement transfers to decoded execution and under which latent and physical optima align. Building on these insights, we propose LAPIS for goal-conditioned latent policy improvement, grounding value-based policy improvement in observed state-latent transitions and using a conditional flow policy to capture their multimodal structure. LAPIS outperforms baselines on most tasks, with the clearest gains in visual manipulation, and remains effective even with only 1% action labels, where competing methods largely collapse.
Source:https://arxiv.org/abs/2611.24482PreviousExtracting Training Data from Diffusion Language Models via InfillingNextThe geometry of neural network decision boundaries
Related posts
Your Series B Is Earmarked for North America. Your Single-Region Backend Is Not Ready.
0dev.to TypeScript
Scheduled posts for a static site without a CMS: a queue folder, a cron slot, a build that can say no
0dev.to JavaScript
How I made a background remover that runs entirely in the browser
0dev.to JavaScript
When "AI Boom" Becomes "Cloud API Regret": The Case for Client-Side Compute
0dev.to JavaScript
Pointing the OpenAI Python and Node SDKs at a model with fewer refusals
0dev.to JavaScript
Optimizing Rust Lock-Free MPMC for Latency Reduction
0dev.to Rust