6arXiv cs.LG (Machine Learning)·
Patch Policy: Lightweight transformer policy using dense ViT patch tokens for robot control
Patch Policy is a new robot learning architecture that enables transformer-based policies to consume dense pre-trained Vision Transformer patch tokens directly, without the overhead of a full vision-language model backbone. A block-causal attention mask preserves temporal causality while allowing the model to attend over many patch tokens per observation. Across four simulated and three real-world environment suites, the method achieves a 40% relative improvement over global-pooled representation baselines and outperforms fine-tuned OpenVLA-OFT by 18% while using roughly 0.7% of its parameters. The work addresses a practical gap between lightweight robot policies and expensive VLA models.
Related guides (2)
Related events (8)
Patch Time Series Transformer in Hugging Face
Hugging Face has integrated PatchTST, a patch-based Transformer architecture for time series forecasting, into its ecosystem. PatchTST applies the patching concept from vision transformers to time series data, dividing sequences into subseries-level patches as input tokens. The blog post covers usage, fine-tuning, and zero-shot transfer capabilities of the model within the Hugging Face Transformers library.
PAC-ACT: Reinforcement learning post-training framework for Action Chunking Transformer robot policies
PAC-ACT is a reinforcement learning post-training framework that fine-tunes pretrained Action Chunking Transformer (ACT) policies for precision industrial contact manipulation tasks. The method reformulates policy optimization at the chunk level, introduces an actor-critic architecture adapted for ACT, and uses a hybrid behavior-prior constraint to prevent distribution shift during online RL fine-tuning. Experiments on industrial contact benchmarks show significant improvements in task success, contact stability, and force safety — including a 46x reduction in force readings above 60 N on a contour-following task — while preserving low latency and GPU memory usage.
HABC: Hierarchical Advantage Weighting for Online RL Fine-Tuning of Vision-Language-Action Policies
Researchers introduce Hierarchical Advantage-Weighted Behavior Cloning (HABC), a method for fine-tuning pretrained Vision-Language-Action (VLA) policies via online RL using only sparse binary episode outcomes. HABC trains separate critic heads for viability and efficiency objectives, combines them via a state-adaptive gate, and applies intervention-aware credit assignment to avoid incorrect supervision across human-intervention boundaries. On three contact-rich bimanual real-robot tasks, HABC improves success rates from SFT baselines of 36%, 44%, and 12% to 92%, 88%, and 38% respectively. The work addresses a fundamental credit assignment problem in robot learning from sparse outcome signals.
TREAD: VLM-based re-labelling framework improves robot policy generalization via dataset augmentation
TREAD (Task Robustness via Re-Labelling Vision-Action Robot Data) is a scalable framework that uses pretrained Vision-Language Models to augment existing robotics datasets without new data collection. The approach decomposes demonstrations into sub-tasks, segments videos accordingly, and generates linguistically diverse instruction labels, enriching language-action pair diversity. Evaluations on the LIBERO benchmark show improved generalization to novel tasks and goals, addressing a key limitation of current robot learning policies.
TempoVLA: Speed-Controllable Vision-Language-Action Policy for Robot Manipulation
Researchers introduce TempoVLA, a Vision-Language-Action model that enables explicit speed control during robot manipulation by conditioning on a speed signal rather than inheriting a fixed speed from training data. The system pairs Variable-Speed Trajectory Augmentation (VSTA), which re-times demonstrations by merging or splitting actions, with a model-side conditioning mechanism. Experiments in simulation and real-world tasks show flexible bidirectional speed control, with dynamic adaptation—accelerating in low-risk transit phases and decelerating for high-risk contact stages—achieved by coupling with a large multimodal model.
CHORUS: Single VLA policy enables decentralized multi-robot collaboration without inter-robot communication
CHORUS is a framework that adapts a single vision-language-action (VLA) backbone to control diverse multi-robot teams in a fully decentralized manner, with each robot running an independent copy conditioned only on its own observations and a robot-identifying prompt. Real-world experiments across tasks like tape measurement, book handovers, and laundry basket lifting show a 64-percentage-point improvement over decentralized from-scratch models and 40-point improvement in reactivity to teammate behavior, while outperforming centralized baselines. The key insight is that pretrained VLA visuomotor priors are sufficient to enable reactive coordination without explicit inter-robot communication or alignment procedures at inference time.
ACPO: Adaptive Clip Policy Optimization improves RLVR training for LLM reasoning
A new arXiv preprint provides theoretical analysis of Reinforcement Learning from Verifiable Rewards (RLVR) updates, identifying off-policy degree and gradient expectation as key factors governing update dynamics. The authors show that differences in gradient steps per rollout substantially affect importance sampling ratio distributions and which tokens dominate updates. Based on this analysis, they propose Adaptive Clip Policy Optimization (ACPO), which adjusts clipping boundaries per token group by empirical variance of importance sampling ratios, outperforming DAPO and CISPO baselines on 3B and 7B models across math, tabular QA, and logic benchmarks.
Ambient Diffusion Policy: imitation learning from suboptimal robot data via noise-dependent co-training
Researchers introduce Ambient Diffusion Policy, a method for robot imitation learning that extracts useful features from suboptimal demonstrations by restricting their contribution to specific diffusion timesteps (high and low noise levels). The approach is grounded in the observation that robot action data follows a spectral power law, inducing global-to-local hierarchy and locality properties in diffusion models. Evaluated across six tasks and four types of suboptimal data, it outperforms co-training baselines by up to 33% when scaled to the Open X-Embodiment dataset.