Searcharxiv⌕ Search

arXiv subjects

Guangyu Zhao

Publications and source records attributed to Guangyu Zhao.

11 recordsLinked to original sources
arXiv2026

STAR: Steering Action Representations during Inference

Vision-language-action models (VLAs) map language instructions and visual observations to robot actions, showing strong capabilities across robotic manipulation tasks. However, in our probing experiment, we find that action representations near the action output do not fully preserve task-relevant visual grounding. Motivated by this late-layer grounding gap, we introduce STAR, for STeering Action Representations, a training-free inference method for pre-trained VLAs. STAR applies an exponential moving average over late-layer residual updates, allowing later action representations to inherit object-aware update directions from earlier layers without model updates or architectural changes. It further uses recent proprioceptive states to adapt the mixing strength, using conservative mixing when execution dynamics change rapidly. Experiments on robot manipulation benchmarks show that STAR improves late-layer object awareness and increases average task success rates by 4.0 percentage points. Our code is available at https://github.com/PKU-VaLuE-Lab/STAR.

cs.RO↗
arXiv2026

MARS: Systematic Multi-Level Acceleration for World Action Models

World Action Models (WAMs) have emerged as a promising alternative for embodied intelligence by jointly predicting future world states and robot actions. However, their heavy visual encoding and iterative generative inference incur substantial latency, limiting their applicability to high-frequency closed-loop control. Existing training-free acceleration methods typically optimize a single stage, yielding limited speedups or degraded control under aggressive reuse. In this work, we systematically profile the inference overhead of WAMs and observe abundant redundancy across Transformer computation, diffusion denoising, and consecutive policy queries. However, the safe removal of such redundancy depends strongly on its impact on action prediction and control sensitivity. Based on these observations, we introduce \textbf{MARS}, a training-free and plug-and-play framework for multi-level WAM acceleration. MARS selectively removes low-impact attention computation, adaptively accelerates diffusion inference according to motion granularity, and speculatively reuses predicted future visual representations across consecutive queries with online verification. Together, these complementary mechanisms provide remarkable acceleration while limiting approximation error. Extensive experiments across LIBERO, LIBERO-PRO, and RoboCasa show that MARS achieves up to $4.00\times$, $3.94\times$, and $2.10\times$ end-to-end speedups, respectively, while maintaining task success rate close to the original policy. Moreover, MARS generalizes across different model backbones and across simulation and real-robot experiments, substantially improving the efficiency--performance trade-off of WAM inference. Our code is available at https://github.com/PKU-VaLuE-Lab/MARS.

cs.CV↗
arXiv2026

EgoSteer: An Open-Source Full-Stack System Towards Steerable Dexterous Manipulation from Egocentric Videos

The enduring vision of general-purpose robots serving humanity hinges fundamentally on policy steerability. However, prevailing paradigms of learning from expert demonstrations demand massive real-world data even on simplified grippers, rendering them prohibitively expensive for high-dimensional, data-scarce dexterous hands. To overcome this bottleneck, we present a full-stack system that scales dexterous VLA pre-training from egocentric human videos and enables data-efficient real-robot post-training. It integrates EgoSmith, a data pipeline that curates in-the-wild egocentric videos into 9,606 hours of pre-training data with 8.3x higher throughput and better accuracy than prior SOTA; a unified Robot Stack for teleoperation and human-in-the-loop correction tailored for dexterous hands; and EgoSteer, a world-model-enhanced VLA operating on a morphology-aligned action space. Human data pre-training equips EgoSteer with language-guided manipulation priors, which are grounded through robot post-training and further refined via DAgger. Empirically, EgoSteer robustly executes free-form instructions across 45 diverse tasks, demonstrating adherence to user intent amid multiple candidate tasks and generalization. The pre-trained model also few-shot adapts to five complex long-horizon tasks, including box folding, on two embodiments with 79% average progress. All system code, datasets, model checkpoints, and an evaluation gallery are publicly available at https://egosteer.github.io/.

cs.RO↗
arXiv2026

Preference Goal Tuning: Post-Training as Latent Control for Frozen Policies

Goal-conditioned policies enable decision-making models to execute diverse behaviors based on specified goals, yet their downstream performance is often highly sensitive to the choice of instructions or prompts. To bypass the limitations of discrete text prompts, we formulate post-training adaptation as a latent control problem, where the goal embedding serves as a continuous control variable to modulate the behavior of a frozen policy. We propose Preference Goal Tuning (PGT), a framework that optimizes this latent control variable to align the induced trajectory distribution with task preferences. Unlike standard fine-tuning that updates policy parameters, PGT keeps the policy frozen and updates only the latent goal using a trajectory-level preference objective. This approach essentially searches for the optimal conditioning input that maximizes the likelihood of preferred behaviors while suppressing undesirable ones. We evaluate PGT on the Minecraft SkillForge benchmark across 17 tasks. With minimal data, PGT achieves average relative improvements of 72.0\% and 81.6\% on two foundation policies, consistently outperforming expert-crafted prompts. Crucially, by decoupling task alignment (latent goal) from physical dynamics (frozen policy), PGT surpasses full fine-tuning by 13.4\% in out-of-distribution settings, demonstrating superior robustness and generalization.

cs.AI↗
arXiv2026

Preserve Support, Not Correspondence: Dynamic Routing for Offline Reinforcement Learning

One-step offline RL actors are attractive because they avoid backpropagating through long iterative samplers and keep inference cheap, but they still have to improve under a critic without drifting away from actions that the dataset can support. In recent one-step extraction pipelines, a strong iterative teacher provides one target action for each latent draw, and the same student output is asked to do both jobs: move toward higher Q and stay near that paired endpoint. If those two directions disagree, the loss resolves them as a compromise on that same sample, even when a nearby better action remains locally supported by the data. We propose DROL, a latent-conditioned one-step actor trained with top-1 dynamic routing. For each state, the actor samples $K$ candidate actions from a bounded latent prior, assigns each dataset action to its nearest candidate, and updates only that winner with Behavior Cloning and critic guidance. Because the routing is recomputed from the current candidate geometry, ownership of a supported region can shift across candidates over the course of learning. This gives a one-step actor room to make local improvements that pointwise extraction struggles to capture, while retaining single-pass inference at test time. On OGBench and D4RL, DROL is competitive with the one-step FQL baseline, improving many OGBench task groups while remaining strong on both AntMaze and Adroit. Project page: https://muzhancun.github.io/preprints/DROL.

cs.LG↗
arXiv2026

System Design for Maintaining Internal State Consistency in Long-Horizon Robotic Tabletop Games

Long-horizon tabletop games pose a distinct systems challenge for robotics: small perceptual or execution errors can invalidate accumulated task state, propagate across decision-making modules, and ultimately derail interaction. This paper studies how to maintain internal state consistency in turn-based, multi-human robotic tabletop games through deliberate system design rather than isolated component improvement. Using Mahjong as a representative long-horizon setting, we present an integrated architecture that explicitly maintains perceptual, execution, and interaction state, partitions high-level semantic reasoning from time-critical perception and control, and incorporates verified action primitives with tactile-triggered recovery to prevent premature state corruption. We further introduce interaction-level monitoring mechanisms to detect turn violations and hidden-information breaches that threaten execution assumptions. Beyond demonstrating complete-game operation, we provide an empirical characterization of failure modes, recovery effectiveness, cross-module error propagation, and hardware-algorithm trade-offs observed during deployment. Our results show that explicit partitioning, monitored state transitions, and recovery mechanisms are critical for sustaining executable consistency over extended play, whereas monolithic or unverified pipelines lead to measurable degradation in end-to-end reliability. The proposed system serves as an empirical platform for studying system-level design principles in long-horizon, turn-based interaction.

cs.RO↗
arXiv2025

When Robots Should Say "I Don't Know": Benchmarking Abstention in Embodied Question Answering

Embodied Question Answering (EQA) requires an agent to interpret language, perceive its environment, and navigate within 3D scenes to produce responses. Existing EQA benchmarks assume that every question must be answered, but embodied agents should know when they do not have sufficient information to answer. In this work, we focus on a minimal requirement for EQA agents, abstention: knowing when to withhold an answer. From an initial study of 500 human queries, we find that 32.4% contain missing or underspecified context. Drawing on this initial study and cognitive theories of human communication errors, we derive five representative categories requiring abstention: actionability limitation, referential underspecification, preference dependence, information unavailability, and false presupposition. We augment OpenEQA by having annotators transform well-posed questions into ambiguous variants outlined by these categories. The resulting dataset, AbstainEQA, comprises 1,636 annotated abstention cases paired with 1,636 original OpenEQA instances for balanced evaluation. Evaluating on AbstainEQA, we find that even the best frontier model only attains 42.79% abstention recall, while humans achieve 91.17%. We also find that scaling, prompting, and reasoning only yield marginal gains, and that fine-tuned models overfit to textual cues. Together, these results position abstention as a fundamental prerequisite for reliable interaction in embodied settings and as a necessary basis for effective clarification.

cs.CV↗
arXiv2024

Whole-Body Control Through Narrow Gaps From Pixels To Action

Flying through body-size narrow gaps in the environment is one of the most challenging moments for an underactuated multirotor. We explore a purely data-driven method to master this flight skill in simulation, where a neural network directly maps pixels and proprioception to continuous low-level control commands. This learned policy enables whole-body control through gaps with different geometries demanding sharp attitude changes (e.g., near-vertical roll angle). The policy is achieved by successive model-free reinforcement learning (RL) and online observation space distillation. The RL policy receives (virtual) point clouds of the gaps' edges for scalable simulation and is then distilled into the high-dimensional pixel space. However, this flight skill is fundamentally expensive to learn by exploring due to restricted feasible solution space. We propose to reset the agent as states on the trajectories by a model-based trajectory optimizer to alleviate this problem. The presented training pipeline is compared with baseline methods, and ablation studies are conducted to identify the key ingredients of our method. The immediate next step is to scale up the variation of gap sizes and geometries in anticipation of emergent policies and demonstrate the sim-to-real transformation.

cs.RO↗
arXiv2024

Learning Speed Adaptation for Flight in Clutter

Animals learn to adapt speed of their movements to their capabilities and the environment they observe. Mobile robots should also demonstrate this ability to trade-off aggressiveness and safety for efficiently accomplishing tasks. The aim of this work is to endow flight vehicles with the ability of speed adaptation in prior unknown and partially observable cluttered environments. We propose a hierarchical learning and planning framework where we utilize both well-established methods of model-based trajectory generation and trial-and-error that comprehensively learns a policy to dynamically configure the speed constraint. Technically, we use online reinforcement learning to obtain the deployable policy. The statistical results in simulation demonstrate the advantages of our method over the constant speed constraint baselines and an alternative method in terms of flight efficiency and safety. In particular, the policy behaves perception awareness, which distinguish it from alternative approaches. By deploying the policy to hardware, we verify that these advantages can be brought to the real world.

cs.RO↗