Reinforcement Learning (RL) can mitigate the causal confusion and distribution shift inherent to imitation learning (IL). However, applying RL to end-to-end autonomous driving (E2E-AD) remains an open problem for its training difficulty, and IL is still the mainstream paradigm in both academia and industry. Recently Model-based Reinforcement Learning (MBRL) have demonstrated promising results in neural planning; however, these methods typically require privileged information as input rather than raw sensor data. We fill this gap by designing Raw2Drive, a dual-stream MBRL approach. Initially, we efficiently train an auxiliary privileged world model paired with a neural planner that uses privileged information as input. Subsequently, we introduce a raw sensor world model trained via our proposed Guidance Mechanism, which ensures consistency between the raw sensor world model and the privileged world model during rollouts. Finally, the raw sensor world model combines the prior knowledge embedded in the heads of the privileged world model to effectively guide the training of the raw sensor policy. Raw2Drive is so far the only RL based end-to-end method on CARLA Leaderboard 2.0, and Bench2Drive and it achieves state-of-the-art performance.

For continuous systems modeled by dynamical equations such as ODEs and SDEs, Bellman's Principle of Optimality takes the form of the Hamilton-Jacobi-Bellman (HJB) equation, which provides the theoretical target of reinforcement learning (RL). Although recent advances in RL successfully leverage this formulation, the existing methods typically assume the underlying dynamics are known a priori because they need explicit access to the coefficient functions of dynamical equations to update the value function following the HJB equation. We address this inherent limitation of HJB-based RL; we propose a model-free approach still targeting the HJB equation and propose the corresponding temporal difference method. We establish exponential convergence of the idealized continuous-time dynamics and empirically demonstrate its potential advantages over transition-kernel-based formulations. The proposed formulation paves the way toward bridging stochastic control and model-free reinforcement learning.

World models predict state transitions in response to actions and are increasingly developed across diverse modalities. However, standard training objectives such as maximum likelihood estimation (MLE) often misalign with task-specific goals of world models, i.e., transition prediction metrics like accuracy or perceptual quality. In this paper, we present RLVR-World, a unified framework that leverages reinforcement learning with verifiable rewards (RLVR) to directly optimize world models for such metrics. Despite formulating world modeling as autoregressive prediction of tokenized sequences, RLVR-World evaluates metrics of decoded predictions as verifiable rewards. We demonstrate substantial performance gains on both language- and video-based world models across domains, including text games, web navigation, and robot manipulation. Our work indicates that, beyond recent advances in reasoning language models, RLVR offers a promising post-training paradigm for enhancing the utility of generative models more broadly. Code, datasets, models, and video samples are available at the project website: https://thuml.github.io/RLVR-World.

Abstract-- Autonomous Surface V ehicles (ASVs) play a crucial role in maritime operations, yet their navigation in shallow-water environments remains challenging due to dynamic disturbances and depth constraints. In this paper, we propose a reinforcement learning (RL) framework for ASV navigation under depth constraints, where the vehicle must reach a target while avoiding unsafe areas with only a single depth measurement per timestep from a downward-facing Single Beam Echosounder (SBES). T o enhance environmental awareness, we integrate Gaussian Process (GP) regression into the RL framework, enabling the agent to progressively estimate a bathymetric depth map from sparse sonar readings. This approach improves decision-making by providing a richer representation of the environment. Furthermore, we demonstrate effective sim-to-real transfer, ensuring that trained policies generalize well to real-world aquatic conditions. Experimental results validate our method's capability to improve ASV navigation performance while maintaining safety in challenging shallow-water environments. I. INTRODUCTION Autonomous Surface V ehicles (ASVs) are unmanned vessels increasingly employed for a variety of maritime operations, including environmental monitoring, search-and-rescue, and surveillance.

Humans exhibit diverse and expressive whole-body movements. However, attaining human-like whole-body coordination in humanoid robots remains challenging, as conventional approaches that mimic whole-body motions often neglect the distinct roles of upper and lower body. This oversight leads to computationally intensive policy learning and frequently causes robot instability and falls during real-world execution. To address these issues, we propose Adversarial Locomotion and Motion Imitation (ALMI), a novel framework that enables adversarial policy learning between upper and lower body. Specifically, the lower body aims to provide robust locomotion capabilities to follow velocity commands while the upper body tracks various motions. Conversely, the upper-body policy ensures effective motion tracking when the robot executes velocity-based movements. Through iterative updates, these policies achieve coordinated whole-body control, which can be extended to loco-manipulation tasks with teleoperation systems. Extensive experiments demonstrate that our method achieves robust locomotion and precise motion tracking in both simulation and on the full-size Unitree H1 robot. Additionally, we release a large-scale whole-body motion control dataset featuring high-quality episodic trajectories from MuJoCo simulations deployable on real robots. The project page is https://almi-humanoid.github.io.

Large reasoning models (LRMs) often consume excessive tokens, inflating computational cost and latency. We challenge the assumption that longer responses improve accuracy. By penalizing reasoning tokens using a discounted reinforcement learning setup (interpretable as a small token cost) and analyzing Blackwell optimality in restricted policy classes, we encourage concise yet accurate reasoning. Experiments confirm our theoretical results that this approach shortens chains of thought while preserving accuracy.

Deep Q Networks (DQN) have shown remarkable success in various reinforcement learning tasks. However, their reliance on associative learning often leads to the acquisition of spurious correlations, hindering their problem-solving capabilities. In this paper, we introduce a novel approach to integrate causal principles into DQNs, leveraging the PEACE (Probabilistic Easy vAriational Causal Effect) formula for estimating causal effects. By incorporating causal reasoning during training, our proposed framework enhances the DQN's understanding of the underlying causal structure of the environment, thereby mitigating the influence of confounding factors and spurious correlations. We demonstrate that integrating DQNs with causal capabilities significantly enhances their problem-solving capabilities without compromising performance. Experimental results on standard benchmark environments showcase that our approach outperforms conventional DQNs, highlighting the effectiveness of causal reasoning in reinforcement learning. Overall, our work presents a promising avenue for advancing the capabilities of deep reinforcement learning agents through principled causal inference.

Autonomous navigation in unstructured environments is essential for field and planetary robotics, where robots must efficiently reach goals while avoiding obstacles under uncertain conditions. Conventional algorithmic approaches often require extensive environment-specific tuning, limiting scalability to new domains. Deep Reinforcement Learning (DRL) provides a data-driven alternative, allowing robots to acquire navigation strategies through direct interactions with their environment. This work investigates the feasibility of DRL policy generalization across visually and topographically distinct simulated domains, where policies are trained in terrestrial settings and validated in a zero-shot manner in extraterrestrial environments. A 3D simulation of an agricultural rover is developed and trained using Proximal Policy Optimization (PPO) to achieve goal-directed navigation and obstacle avoidance in farmland settings. The learned policy is then evaluated in a lunar-like simulated environment to assess transfer performance. The results indicate that policies trained under terrestrial conditions retain a high level of effectiveness, achieving close to 50\% success in lunar simulations without the need for additional training and fine-tuning. This underscores the potential of cross-domain DRL-based policy transfer as a promising approach to developing adaptable and efficient autonomous navigation for future planetary exploration missions, with the added benefit of minimizing retraining costs.

CNOT gates are fundamental to quantum computing, as they facilitate entanglement, a crucial resource for quantum algorithms. Certain classes of quantum circuits are constructed exclusively from CNOT gates. Given their widespread use, it is imperative to minimise the number of CNOT gates employed. This problem, known as CNOT minimisation, remains an open challenge, with its computational complexity yet to be fully characterised. In this work, we introduce a novel reinforcement learning approach to address this task. Instead of training multiple reinforcement learning agents for different circuit sizes, we use a single agent up to a fixed size $m$. Matrices of sizes different from m are preprocessed using either embedding or Gaussian striping. To assess the efficacy of our approach, we trained an agent with m = 8, and evaluated it on matrices of size n that range from 3 to 15. The results we obtained show that our method overperforms the state-of-the-art algorithm as the value of n increases.

Cross-domain offline reinforcement learning (RL) seeks to enhance sample efficiency in offline RL by utilizing additional offline source datasets. A key challenge is to identify and utilize source samples that are most relevant to the target domain. Existing approaches address this challenge by measuring domain gaps through domain classifiers, target transition dynamics modeling, or mutual information estimation using contrastive loss. However, these methods often require large target datasets, which is impractical in many real-world scenarios. In this work, we address cross-domain offline RL under a limited target data setting, identifying two primary challenges: (1) Dataset imbalance, which is caused by large source and small target datasets and leads to overfitting in neural network-based domain gap estimators, resulting in uninformative measurements; and (2) Partial domain overlap, where only a subset of the source data is closely aligned with the target domain. To overcome these issues, we propose DmC, a novel framework for cross-domain offline RL with limited target samples. Specifically, DmC utilizes $k$-nearest neighbor ($k$-NN) based estimation to measure domain proximity without neural network training, effectively mitigating overfitting. Then, by utilizing this domain proximity, we introduce a nearest-neighbor-guided diffusion model to generate additional source samples that are better aligned with the target domain, thus enhancing policy learning with more effective source samples. Through theoretical analysis and extensive experiments in diverse MuJoCo environments, we demonstrate that DmC significantly outperforms state-of-the-art cross-domain offline RL methods, achieving substantial performance gains.