Wenshuai Zhao

Yi Zhao, Wenshuai Zhao, Rinu Boney et al.

Reinforcement learning is able to solve complex sequential decision-making tasks but is currently limited by sample efficiency and required computation. To improve sample efficiency, recent work focuses on model-based RL which interleaves model learning with planning. Recent methods further utilize policy learning, value estimation, and, self-supervised learning as auxiliary objectives. In this paper we show that, surprisingly, a simple representation learning approach relying only on a latent dynamics model trained by latent temporal consistency is sufficient for high-performance RL. This applies when using pure planning with a dynamics model conditioned on the representation, but, also when utilizing the representation as policy and value function features in model-free RL. In experiments, our approach learns an accurate dynamics model to solve challenging high-dimensional locomotion tasks with online planners while being 4.1 times faster to train compared to ensemble-based methods. With model-free RL without planning, especially on high-dimensional tasks, such as the DeepMind Control Suite Humanoid and Dog tasks, our approach outperforms model-free methods by a large margin and matches model-based methods' sample efficiency while training 2.4 times faster.

Wenshuai Zhao, Yi Zhao, Zhiyuan Li et al.

*Relative overgeneralization* (RO) occurs in cooperative multi-agent learning tasks when agents converge towards a suboptimal joint policy due to overfitting to suboptimal behavior of other agents. No methods have been proposed for addressing RO in multi-agent policy gradient (MAPG) methods although these methods produce state-of-the-art results. To address this gap, we propose a general, yet simple, framework to enable optimistic updates in MAPG methods that alleviate the RO problem. Our approach involves clipping the advantage to eliminate negative values, thereby facilitating optimistic updates in MAPG. The optimism prevents individual agents from quickly converging to a local optimum. Additionally, we provide a formal analysis to show that the proposed method retains optimality at a fixed point. In extensive evaluations on a diverse set of tasks including the *Multi-agent MuJoCo* and *Overcooked* benchmarks, our method outperforms strong baselines on 13 out of 19 tested tasks and matches the performance on the rest.

Jiarui Guan, Wenshuai Zhao, Zhengtao Zou et al.

Video variational autoencoders (VAEs) used in latent diffusion models typically require a sufficiently large number of latent channels to ensure high-quality video reconstruction. However, recent studies have revealed that an excessive number of latent channels can impede the convergence of latent diffusion models and deteriorate their generative performance, even when reconstruction quality remains high. We propose a latent compression method that removes high-frequency components in video latent representations rather than directly reducing the number of channels, which often compromises reconstruction fidelity. Experimental results demonstrate that the proposed method achieves superior video reconstruction quality compared to strong baselines while maintaining the same overall compression ratio.

Yihao Wang, Yang Miao, Wenshuai Zhao et al.

Articulation perception aims to recover the motion and structure of articulated objects (e.g., drawers and cupboards), and is fundamental to 3D scene understanding in robotics, simulation, and animation. Existing learning-based methods rely heavily on supervised training with high-quality 3D data and manual annotations, limiting scalability and diversity. To address this limitation, we propose PAWS, a method that directly extracts object articulations from hand-object interactions in large-scale in-the-wild egocentric videos. We evaluate our method on the public data sets, including HD-EPIC and Arti4D data sets, achieving significant improvements over baselines. We further demonstrate that the extracted articulations benefit downstream tasks, including fine-tuning 3D articulation prediction models and enabling robot manipulation. See the project website at https://aaltoml.github.io/PAWS/.

Jiarui Guan, Wenshuai Zhao, Yue Pei et al.

Robot policy learning benefits from world-action models that capture environment dynamics, but pixel-level prediction entangles dynamics with nuisance factors such as lighting and texture, making learned representations vulnerable to task-irrelevant visual variation. We propose JOPAT, a JOint Pixel-And-Track World-Action Model that predicts latent visual observations, 2D point tracks with visibility, and actions in a single denoising diffusion transformer. The key insight is that tracks provide an explicit representation of motion that captures long-horizon dynamics and remains robust under occlusion or partial out-of-frame motion, offering greater utility than modeling pixel appearance alone. On LIBERO and real-world LeRobot tasks, JOPAT improves over pixel-based baselines, with the largest gains on long-horizon tasks involving occlusion, object interaction, and off-screen motion.

Wenshuai Zhao, Zhiyuan Li, Joni Pajarinen

The number of agents can be an effective curriculum variable for controlling the difficulty of multi-agent reinforcement learning (MARL) tasks. Existing work typically uses manually defined curricula such as linear schemes. We identify two potential flaws while applying existing reward-based automatic curriculum learning methods in MARL: (1) The expected episode return used to measure task difficulty has high variance; (2) Credit assignment difficulty can be exacerbated in tasks where increasing the number of agents yields higher returns which is common in many MARL tasks. To address these issues, we propose to control the curriculum by using a TD-error based *learning progress* measure and by letting the curriculum proceed from an initial context distribution to the final task specific one. Since our approach maintains a distribution over the number of agents and measures learning progress rather than absolute performance, which often increases with the number of agents, we alleviate problem (2). Moreover, the learning progress measure naturally alleviates problem (1) by aggregating returns. In three challenging sparse-reward MARL benchmarks, our approach outperforms state-of-the-art baselines.

Zhiyuan Li, Rongzhen Zhao, Wenyan Yang et al.

The de facto approach in video object-centric learning maintains temporal consistency through learned dynamics modules that predict future object representations, called slots. We demonstrate that these predictors function as expensive approximations of discrete correspondence problems. Modern self-supervised vision backbones already encode instance-discriminative features that distinguish objects reliably. Exploiting these features eliminates the need for learned temporal prediction. We introduce Grounded Correspondence, a framework that replaces learned transition functions with deterministic bipartite matching. Slots initialize from salient regions in frozen backbone features. Frame-to-frame identity is maintained through Hungarian matching on slot representations. The approach requires zero learnable parameters for temporal modeling yet achieves competitive performance on MOVi-D, MOVi-E, and YouTube-VIS. Project page: https://magenta-sherbet-85b101.netlify.app/

Zhiyuan Li, Wenyan Yang, Wenshuai Zhao et al.

Learning robotic manipulation from human videos is a promising solution to the data bottleneck in robotics, but the distribution shift between humans and robots remains a critical challenge. Existing approaches often produce entangled representations, where task-relevant information is coupled with human-specific kinematics, limiting their adaptability. We propose a generative framework for cross-embodiment video editing that directly addresses this by learning explicitly disentangled task and embodiment representations. Our method factorizes a demonstration video into two orthogonal latent spaces by enforcing a dual contrastive objective: it minimizes mutual information between the spaces to ensure independence while maximizing intra-space consistency to create stable representations. A parameter-efficient adapter injects these latent codes into a frozen video diffusion model, enabling the synthesis of a coherent robot execution video from a single human demonstration, without requiring paired cross-embodiment data. Experiments show our approach generates temporally consistent and morphologically accurate robot demonstrations, offering a scalable solution to leverage internet-scale human video for robot learning.

Yi Zhao, Aidan Scannell, Wenshuai Zhao et al.

Leveraging offline data is a promising way to improve the sample efficiency of online reinforcement learning (RL). This paper expands the pool of usable data for offline-to-online RL by leveraging abundant non-curated data that is reward-free, of mixed quality, and collected across multiple embodiments. Although learning a world model appears promising for utilizing such data, we find that naive fine-tuning fails to accelerate RL training on many tasks. Through careful investigation, we attribute this failure to the distributional shift between offline and online data during fine-tuning. To address this issue and effectively use the offline data, we propose two essential techniques: \emph{i)} experience rehearsal and \emph{ii)} execution guidance. With these modifications, the non-curated offline data substantially improves RL's sample efficiency. Under limited sample budgets, our method achieves a 102.8\% relative improvement in aggregate score over learning-from-scratch baselines across 72 visuomotor tasks spanning 6 embodiments. On challenging tasks such as locomotion and robotic manipulation, it outperforms prior methods that utilize offline data by a decent margin.

Wenshuai Zhao, Zhiyuan Li, Yi Zhao et al.

Diffusion models have shown remarkable success across a wide range of generative tasks. However, they often suffer from spatially inconsistent generation, arguably due to the inherent locality of their denoising mechanisms. This can yield samples that are locally plausible but globally inconsistent. To mitigate this issue, we propose sparsely supervised learning for diffusion models, a simple yet effective masking strategy that can be implemented with only a few lines of code. Interestingly, the experiments show that it is safe to mask up to 98\% of pixels during diffusion model training. Our method delivers competitive FID scores across experiments and, most importantly, avoids training instability on small datasets. Moreover, the masking strategy reduces memorization and promotes the use of essential contextual information during generation.

Zhiyuan Li, Wenshuai Zhao, Joni Pajarinen

Despite much progress in training distributed artificial intelligence (AI), building cooperative multi-agent systems with multi-agent reinforcement learning (MARL) faces challenges in sample efficiency, interpretability, and transferability. Unlike traditional learning-based methods that require extensive interaction with the environment, large language models (LLMs) demonstrate remarkable capabilities in zero-shot planning and complex reasoning. However, existing LLM-based approaches heavily rely on text-based observations and struggle with the non-Markovian nature of multi-agent interactions under partial observability. We present COMPASS, a novel multi-agent architecture that integrates vision-language models (VLMs) with a dynamic skill library and structured communication for decentralized closed-loop decision-making. The skill library, bootstrapped from demonstrations, evolves via planner-guided tasks to enable adaptive strategies. COMPASS propagates entity information through multi-hop communication under partial observability. Evaluations on the improved StarCraft Multi-Agent Challenge (SMACv2) demonstrate COMPASS's strong performance against state-of-the-art MARL baselines across both symmetric and asymmetric scenarios. Notably, in the symmetric Protoss 5v5 task, COMPASS achieved a 57\% win rate, representing a 30 percentage point advantage over QMIX (27\%). Project page can be found at https://stellar-entremet-1720bb.netlify.app/.

Zhiyuan Li, Wenshuai Zhao, Lijun Wu et al.

In multi-agent reinforcement learning, centralized training with decentralized execution (CTDE) methods typically assume that agents make decisions based on their local observations independently, which may not lead to a correlated joint policy with coordination. Coordination can be explicitly encouraged during training and individual policies can be trained to imitate the correlated joint policy. However, this may lead to an \textit{asymmetric learning failure} due to the observation mismatch between the joint and individual policies. Inspired by the concept of correlated equilibrium, we introduce a \textit{strategy modification} called AgentMixer that allows agents to correlate their policies. AgentMixer combines individual partially observable policies into a joint fully observable policy non-linearly. To enable decentralized execution, we introduce \textit{Individual-Global-Consistency} to guarantee mode consistency during joint training of the centralized and decentralized policies and prove that AgentMixer converges to an $ε$-approximate Correlated Equilibrium. In the Multi-Agent MuJoCo, SMAC-v2, Matrix Game, and Predator-Prey benchmarks, AgentMixer outperforms or matches state-of-the-art methods.

Wenshuai Zhao, Jorge Peña Queralta, Tomi Westerlund

Deep reinforcement learning has recently seen huge success across multiple areas in the robotics domain. Owing to the limitations of gathering real-world data, i.e., sample inefficiency and the cost of collecting it, simulation environments are utilized for training the different agents. This not only aids in providing a potentially infinite data source, but also alleviates safety concerns with real robots. Nonetheless, the gap between the simulated and real worlds degrades the performance of the policies once the models are transferred into real robots. Multiple research efforts are therefore now being directed towards closing this sim-to-real gap and accomplish more efficient policy transfer. Recent years have seen the emergence of multiple methods applicable to different domains, but there is a lack, to the best of our knowledge, of a comprehensive review summarizing and putting into context the different methods. In this survey paper, we cover the fundamental background behind sim-to-real transfer in deep reinforcement learning and overview the main methods being utilized at the moment: domain randomization, domain adaptation, imitation learning, meta-learning and knowledge distillation. We categorize some of the most relevant recent works, and outline the main application scenarios. Finally, we discuss the main opportunities and challenges of the different approaches and point to the most promising directions.

Wenshuai Zhao, Jorge Peña Queralta, Li Qingqing et al.

Current research directions in deep reinforcement learning include bridging the simulation-reality gap, improving sample efficiency of experiences in distributed multi-agent reinforcement learning, together with the development of robust methods against adversarial agents in distributed learning, among many others. In this work, we are particularly interested in analyzing how multi-agent reinforcement learning can bridge the gap to reality in distributed multi-robot systems where the operation of the different robots is not necessarily homogeneous. These variations can happen due to sensing mismatches, inherent errors in terms of calibration of the mechanical joints, or simple differences in accuracy. While our results are simulation-based, we introduce the effect of sensing, calibration, and accuracy mismatches in distributed reinforcement learning with proximal policy optimization (PPO). We discuss on how both the different types of perturbances and how the number of agents experiencing those perturbances affect the collaborative learning effort. The simulations are carried out using a Kuka arm model in the Bullet physics engine. This is, to the best of our knowledge, the first work exploring the limitations of PPO in multi-robot systems when considering that different robots might be exposed to different environments where their sensors or actuators have induced errors. With the conclusions of this work, we set the initial point for future work on designing and developing methods to achieve robust reinforcement learning on the presence of real-world perturbances that might differ within a multi-robot system.

Wenshuai Zhao, Jorge Peña Queralta, Li Qingqing et al.

The integration of edge computing in next-generation mobile networks is bringing low-latency and high-bandwidth ubiquitous connectivity to a myriad of cyber-physical systems. This will further boost the increasing intelligence that is being embedded at the edge in various types of autonomous systems, where collaborative machine learning has the potential to play a significant role. This paper discusses some of the challenges in multi-agent distributed deep reinforcement learning that can occur in the presence of byzantine or malfunctioning agents. As the simulation-to-reality gap gets bridged, the probability of malfunctions or errors must be taken into account. We show how wrong discrete actions can significantly affect the collaborative learning effort. In particular, we analyze the effect of having a fraction of agents that might perform the wrong action with a given probability. We study the ability of the system to converge towards a common working policy through the collaborative learning process based on the number of experiences from each of the agents to be aggregated for each policy update, together with the fraction of wrong actions from agents experiencing malfunctions. Our experiments are carried out in a simulation environment using the Atari testbed for the discrete action spaces, and advantage actor-critic (A2C) for the distributed multi-agent training.

Wenshuai Zhao, Dihong Jiang, Jorge Peña Queralta et al.

Accurate segmentation of kidneys and kidney tumors is an essential step for radiomic analysis as well as developing advanced surgical planning techniques. In clinical analysis, the segmentation is currently performed by clinicians from the visual inspection images gathered through a computed tomography (CT) scan. This process is laborious and its success significantly depends on previous experience. Moreover, the uncertainty in the tumor location and heterogeneity of scans across patients increases the error rate. To tackle this issue, computer-aided segmentation based on deep learning techniques have become increasingly popular. We present a multi-scale supervised 3D U-Net, MSS U-Net, to automatically segment kidneys and kidney tumors from CT images. Our architecture combines deep supervision with exponential logarithmic loss to increase the 3D U-Net training efficiency. Furthermore, we introduce a connected-component based post processing method to enhance the performance of the overall process. This architecture shows superior performance compared to state-of-the-art works using data from KiTS19 public dataset, with the Dice coefficient of kidney and tumor up to 0.969 and 0.805 respectively. The segmentation techniques introduced in this paper have been tested in the KiTS19 challenge with its corresponding dataset.

Wenshuai Zhao, Zengfeng Zeng

U-Net has achieved huge success in various medical image segmentation challenges. Kinds of new architectures with bells and whistles might succeed in certain dataset when employed with optimal hyper-parameter, but their generalization always can't be guaranteed. Here, we focused on the basic U-Net architecture and proposed a multi scale supervised 3D U-Net for the segmentation task in KiTS19 challenge. To enhance the performance, our work can be summarized as three folds: first, we used multi scale supervision in the decoder pathway, which could encourage the network to predict right results from the deep layers; second, with the aim to alleviate the bad effect from the sample imbalance of kidney and tumor, we adopted exponential logarithmic loss; third, a connected-component based post processing method was designed to remove the obviously wrong voxels. In the published KiTS19 training dataset (totally 210 patients), we divided 42 patients to be test dataset and finally obtained DICE scores of 0.969 and 0.805 for the kidney and tumor respectively. In the challenge, we finally achieved the 7th place among 106 teams with the Composite Dice of 0.8961, namely 0.9741 for kidney and 0.8181 for tumor.