Reinforcement Learning (RL) applications in real-world scenarios must prioritize safety and reliability, which impose strict constraints on agent behavior. Model-based RL leverages predictive world models for action planning and policy optimization, but inherent model inaccuracies can lead to catastrophic failures in safety-critical settings. We propose a novel model-based RL framework that jointly optimizes task performance and safety. To address world model errors, our method incorporates an adaptive mechanism that dynamically switches between model-based planning and direct policy execution. We resolve the objective mismatch problem of traditional model-based approaches using an implicit world model. Furthermore, our framework employs dynamic safety thresholds that adapt to the agent's evolving capabilities, consistently selecting actions that surpass safe policy suggestions in both performance and safety. Experiments demonstrate significant improvements over non-adaptive methods, showing that our approach optimizes safety and performance simultaneously rather than merely meeting minimum safety requirements. The proposed framework achieves robust performance on diverse safety-critical continuous control tasks, outperforming existing methods.

--We introduce Unsupervised Meta-T esting with Conditional Neural Processes (UMCNP), a novel hybrid few-shot meta-reinforcement learning (meta-RL) method that uniquely combines, yet distinctly separates, parameterized policy gradient-based (PPG) and task inference-based few-shot meta-RL. T ai-lored for settings where the reward signal is missing during meta-testing, our method increases sample efficiency without requiring additional samples in meta-training. UMCNP leverages the efficiency and scalability of Conditional Neural Processes (CNPs) to reduce the number of online interactions required in meta-testing. During meta-training, samples previously collected through PPG meta-RL are efficiently reused for learning task inference in an offline manner . This approach allows us to generate rollouts for self-adaptation by interacting with the learned dynamics model. We demonstrate our method can adapt to an unseen test task using significantly fewer samples during meta-testing than the baselines in 2D-Point Agent and continuous control meta-RL benchmarks, namely, cartpole with unknown angle sensor bias, walker agent with randomized dynamics parameters. ESPITE the remarkable achievements of deep reinforcement learning (RL) algorithms [1], [2], they are hindered by poor sample efficiency and limited generalizability. Few-shot meta-reinforcement learning (meta-RL) framework [3], [4] aims to overcome these limitations by learning a distribution of tasks during meta-training and adapting to a new task from the same distribution using a few samples from the new task in meta-testing.

We propose a novel composite reward function for a reinforcement learning (RL) trading agent that explicitly balances return and risk by combining four differentiable components--annualized return, downside risk, differential return, and the Treynor ratio. Unlike traditional single-metric objectives (e.g., Sharpe or cumulative return), which can encourage reward hacking or over-optimization of one aspect of trading, our formulation is inherently modular and weighted w

Autonomous driving promises significant advancements in mobility, road safety and traffic efficiency, yet reinforcement learning and imitation learning face safe-exploration and distribution-shift challenges. Although human-AI collaboration alleviates these issues, it often relies heavily on extensive human intervention, which increases costs and reduces efficiency. This paper develops a confidence-guided human-AI collaboration (C-HAC) strategy to overcome these limitations. First, C-HAC employs a distributional proxy value propagation method within the distributional soft actor-critic (DSAC) framework. By leveraging return distributions to represent human intentions C-HAC achieves rapid and stable learning of human-guided policies with minimal human interaction. Subsequently, a shared control mechanism is activated to integrate the learned human-guided policy with a self-learning policy that maximizes cumulative rewards. This enables the agent to explore independently and continuously enhance its performance beyond human guidance. Finally, a policy confidence evaluation algorithm capitalizes on DSAC's return distribution networks to facilitate dynamic switching between human-guided and self-learning policies via a confidence-based intervention function. This ensures the agent can pursue optimal policies while maintaining safety and performance guarantees. Extensive experiments across diverse driving scenarios reveal that C-HAC significantly outperforms conventional methods in terms of safety, efficiency, and overall performance, achieving state-of-the-art results. The effectiveness of the proposed method is further validated through real-world road tests in complex traffic conditions. The videos and code are available at: https://github.com/lzqw/C-HAC.

Model-free reinforcement learning has emerged as a powerful method for developing robust robot control policies capable of navigating through complex and unstructured environments. The effectiveness of these methods hinges on two essential elements: (1) the use of massively parallel physics simulations to expedite policy training, and (2) an environment generator tasked with crafting sufficiently challenging yet attainable environments to facilitate continuous policy improvement. Existing methods of outdoor environment generation often rely on heuristics constrained by a set of parameters, limiting the diversity and realism. In this work, we introduce ADEPT, a novel \textbf{A}daptive \textbf{D}iffusion \textbf{E}nvironment for \textbf{P}olicy \textbf{T}ransfer in the zero-shot sim-to-real fashion that leverages Denoising Diffusion Probabilistic Models to dynamically expand existing training environments by adding more diverse and complex environments adaptive to the current policy. ADEPT guides the diffusion model's generation process through initial noise optimization, blending noise-corrupted environments from existing training environments weighted by the policy's performance in each corresponding environment. By manipulating the noise corruption level, ADEPT seamlessly transitions between generating similar environments for policy fine-tuning and novel ones to expand training diversity. To benchmark ADEPT in off-road navigation, we propose a fast and effective multi-layer map representation for wild environment generation. Our experiments show that the policy trained by ADEPT outperforms both procedural generated and natural environments, along with popular navigation methods.

Real-world datasets collected from sensors or human inputs are prone to noise and errors, posing significant challenges for applying offline reinforcement learning (RL). While existing methods have made progress in addressing corrupted actions and rewards, they remain insufficient for handling corruption in high-dimensional state spaces and for cases where multiple elements in the dataset are corrupted simultaneously. Diffusion models, known for their strong denoising capabilities, offer a promising direction for this problem-but their tendency to overfit noisy samples limits their direct applicability. To overcome this, we propose Ambient Diffusion-Guided Dataset Recovery (ADG), a novel approach that pioneers the use of diffusion models to tackle data corruption in offline RL. First, we introduce Ambient Denoising Diffusion Probabilistic Models (DDPM) from approximated distributions, which enable learning on partially corrupted datasets with theoretical guarantees. Second, we use the noise-prediction property of Ambient DDPM to distinguish between clean and corrupted data, and then use the clean subset to train a standard DDPM. Third, we employ the trained standard DDPM to refine the previously identified corrupted data, enhancing data quality for subsequent offline RL training. A notable strength of ADG is its versatility-it can be seamlessly integrated with any offline RL algorithm. Experiments on a range of benchmarks, including MuJoCo, Kitchen, and Adroit, demonstrate that ADG effectively mitigates the impact of corrupted data and improves the robustness of offline RL under various noise settings, achieving state-of-the-art results.

Despite recent progress in large-scale reinforcement learning (RL) for reasoning, the training recipe for building high-performing reasoning models remains elusive. Key implementation details of frontier models, such as DeepSeek-R1, including data curation strategies and RL training recipe, are often omitted. Moreover, recent research indicates distillation remains more effective than RL for smaller models. In this work, we demonstrate that large-scale RL can significantly enhance the reasoning capabilities of strong, small- and mid-sized models, achieving results that surpass those of state-of-the-art distillation-based models. We systematically study the RL training process through extensive ablations and propose a simple yet effective approach: first training on math-only prompts, then on code-only prompts. Notably, we find that math-only RL not only significantly enhances the performance of strong distilled models on math benchmarks (e.g., +14.6% / +17.2% on AIME 2025 for the 7B / 14B models), but also code reasoning tasks (e.g., +6.8% / +5.8% on LiveCodeBench for the 7B / 14B models). In addition, extended code-only RL iterations further improve performance on code benchmarks with minimal or no degradation in math results. We develop a robust data curation pipeline to collect challenging prompts with high-quality, verifiable answers and test cases to enable verification-based RL across both domains. Finally, we identify key experimental insights, including curriculum learning with progressively increasing response lengths and the stabilizing effect of on-policy parameter updates. We find that RL not only elicits the foundational reasoning capabilities acquired during pretraining and supervised fine-tuning (e.g., distillation), but also pushes the limits of the model's reasoning ability, enabling it to solve problems that were previously unsolvable.

Algorithms designed for routing problems typically rely on high-quality candidate edges to guide their search, aiming to reduce the search space and enhance the search efficiency. However, many existing algorithms, like the classical Lin-Kernighan-Helsgaun (LKH) algorithm for the Traveling Salesman Problem (TSP), often use predetermined candidate edges that remain static throughout local searches. This rigidity could cause the algorithm to get trapped in local optima, limiting its potential to find better solutions. To address this issue, we propose expanding the candidate sets to include other promising edges, providing them an opportunity for selection. Specifically, we incorporate multi-armed bandit models to dynamically select the most suitable candidate edges in each iteration, enabling LKH to make smarter choices and lead to improved solutions. Extensive experiments on multiple TSP benchmarks show the excellent performance of our method. Moreover, we employ this bandit-based method to LKH-3, an extension of LKH tailored for solving various TSP variant problems, and our method also significantly enhances LKH-3's performance across typical TSP variants.

Offline cooperative multi-agent reinforcement learning (MARL) faces unique challenges due to distributional shifts, particularly stemming from the high dimensionality of joint action spaces and the presence of out-of-distribution joint action selections. In this work, we highlight that a fundamental challenge in offline MARL arises from the multi-equilibrium nature of cooperative tasks, which induces a highly multimodal joint behavior policy space coupled with heterogeneous-quality behavior data. This makes it difficult for individual policy regularization to align with a consistent coordination pattern, leading to the policy distribution shift problems. To tackle this challenge, we design a sequential score function decomposition method that distills per-agent regularization signals from the joint behavior policy, which induces coordinated modality selection under decentralized execution constraints. Then we leverage a flexible diffusion-based generative model to learn these score functions from multimodal offline data, and integrate them into joint-action critics to guide policy updates toward high-reward, in-distribution regions under a shared team reward. Our approach achieves state-of-the-art performance across multiple particle environments and Multi-agent MuJoCo benchmarks consistently. To the best of our knowledge, this is the first work to explicitly address the distributional gap between offline and online MARL, paving the way for more generalizable offline policy-based MARL methods.

-- Human motion retargeting for humanoid robots, transferring human motion data to robots for imitation, presents significant challenges but offers considerable potential for real-world applications. Traditionally, this process relies on human demonstrations captured through pose estimation or motion capture systems. In this paper, we explore a text-driven approach to mapping human motion to humanoids. T o address the inherent discrepancies between the generated motion representations and the kinematic constraints of humanoid robots, we propose an angle signal network based on norm-position and rotation loss (NPR Loss). It generates joint angles, which serve as inputs to a reinforcement learning-based whole-body joint motion control policy. The policy ensures tracking of the generated motions while maintaining the robot's stability during execution. Our experimental results demonstrate the efficacy of this approach, successfully transferring text-driven human motion to a real humanoid robot NAO. Humanoid robots have long been recognized for their potential to mimic human actions due to their anthropomorphic structure.