arXiv subjects

Yuquan Xue

Publications and source records attributed to Yuquan Xue.

3 recordsLinked to original sources

arXiv2026

WorldSample: Closed-loop Real-robot RL with World Modelling

Reinforcement learning (RL) can overcome the demonstration-coverage limitation of imitation learning (IL) by allowing robots to improve through trial-and-error interaction beyond the states observed in demonstrations. However, deploying RL on real robots remains constrained by high interaction costs, since each physical rollout is costly and reflects only one realized action-outcome path. To address this challenge, we propose WorldSample, a physically grounded data augmentation framework for real-robot RL that closes a real-synthetic loop between physical rollouts, world-model generation, and policy improvement. Grounded on real rollouts, WorldSample generates high-fidelity synthetic transitions through a post-trained world model, which greatly lowers the visual hallucination. Specifically, rather than simply using these transitions as real-world experience, WorldSample introduces Policy-Paced Learning (PPL) to regulate the training process through sample selection and scheduling, balancing useful augmentation against value overestimation and mitigating the hallucination-induced noise. Experiments on robot manipulation tasks involving contact-rich and precise tasks show that WorldSample improves policy success rate by 28% while reducing training steps by 59% compared with baselines. Furthermore, WorldSample improves world model visual fidelity by 19.4dB in PSNR and 0.47 in SSIM over demonstration-only post-training, validating the effectiveness of the real-synthetic loop for both policy and world model performance.

cs.RO↗

arXiv2026

Preference-Calibrated Human-in-the-Loop Reinforcement Learning for Robotic Manipulation

Human-in-the-loop reinforcement learning (HIL-RL) improves sample efficiency in real-robot manipulation through online human intervention. However, successful trajectories may include suboptimal actions that deviate from the desired task-execution path and force human intervention. Existing HIL-RL methods typically apply the consistent credit assignment principle to all transitions, uniformly propagating discounted terminal rewards through suboptimal segments, ignoring the actual contribution of each transition to task success. This overestimates Q-values for critic learning and indirectly misguides actor updates toward suboptimal behavior patterns. To this end, we propose PACT, a Preference-calibrated Actor-Critic Training framework that leverages the implicit preference signals induced by intervention to perform credit reassignment on identified suboptimal segments while directly guiding policy training for unbiased critic-actor learning. Specifically, we first design a progress model that learns from human demonstration and identifies suboptimal segments for credit correction. Then, from the human action and resampled policy action at the intervention state, we build preference pairs to define a counterfactual advantage that penalizes Bellman targets of the identified suboptimal segment, enabling directional credit calibration. Moreover, we directly align the policy with human corrective actions in the bounded mean space, providing an additional signal beyond critic-guided updates. Across five real-robot manipulation tasks, PACT improves the average success rate by 24.5% and achieves 1.3 times faster convergence, thereby improving both RL sample efficiency and performance. Code is available at https://anonymous.4open.science/r/HILRL-A1X-BC05.

cs.RO↗

arXiv2025

RESample: A Robust Data Augmentation Framework via Exploratory Sampling for Robotic Manipulation

Vision-Language-Action (VLA) models have shown strong manipulation capability when trained with large-scale imitation learning datasets. However, these datasets that predominantly consist of successful trajectories rarely provide the corrective supervision required when execution deviates from standard demonstrations. During deployment, these physical deviations lead to the distributional shift that drives policy to failure scenarios, yet missing failure recovery data prevents policy from correcting these execution deviations. To address this failure recovery problem, we propose a coverage-guided data augmentation framework RESample to actively supplement demonstration datasets for failure recovery. Specifically, to guide the augmentation and generate failure modes that possibly appear in the real world, RESample trains a conservative coverage function to identify failure cases that reside within the actual data distribution but are missing in the standard successful demonstrations. Guided by the evaluated coverage discrepancy, we perform exploratory sampling to actively sample exploration behaviors followed by recovery actions, extending the coverage of training data with failure recovery trajectories. With the augmented trajectory, the refined policy, which deviated in real settings, can recover from failure. Experiments on the LIBERO benchmark and real-world manipulation tasks show that RESample consistently improves policy success rates, achieving up to 12% absolute gain with no more than 20% additional samples.

cs.RO↗