Context-based offline meta-reinforcement learning (OMRL) methods have achieved appealing success by leveraging pre-collected offline datasets to develop task representations that guide policy learning. However, current context-based OMRL methods often introduce spurious correlations, where task components are incorrectly correlated due to confounders. These correlations can degrade policy performance when the confounders in the test task differ from those in the training task. To address this problem, we propose CausalCOMRL, a context-based OMRL method that integrates causal representation learning. This approach uncovers causal relationships among the task components and incorporates the causal relationships into task representations, enhancing the generalizability of RL agents. We further improve the distinction of task representations from different tasks by using mutual information optimization and contrastive learning. Utilizing these causal task representations, we employ SAC to optimize policies on meta-RL benchmarks. Experimental results show that CausalCOMRL achieves better performance than other methods on most benchmarks.

Large language models (LLMs) have recently been employed as agents to solve sequential decision-making tasks such as Bayesian optimization and multi-armed bandits (MAB). These works usually adopt an LLM for sequential action selection by providing it with a fixed, manually designed meta-prompt. However, numerous previous works have found that the prompt has a significant impact on the performance of the LLM, which calls for a method to automatically optimize the meta-prompt for LLM-based agents. Unfortunately, the non-stationarity in the reward observations during LLM-based sequential decision-making makes meta-prompt optimization highly challenging. To address this challenge, we draw inspirations from adversarial bandit algorithms, which are inherently capable of handling non-stationary reward observations. Building on this foundation, we propose our EXPonential-weight algorithm for prompt Optimization} (EXPO) to automatically optimize the task description and meta-instruction in the meta-prompt for LLM-based agents. We also extend EXPO to additionally optimize the exemplars (i.e., history of interactions) in the meta-prompt to further enhance the performance, hence introducing our EXPO-ES algorithm. We use extensive experiments to show that our algorithms significantly improve the performance of LLM-based sequential decision-making.

Post-processing networks (PPNs) are components that modify the outputs of arbitrary modules in task-oriented dialogue systems and are optimized using reinforcement learning (RL) to improve the overall task completion capability of the system. However, previous PPN-based approaches have been limited to handling only a subset of modules within a system, which poses a significant limitation in improving the system performance. In this study, we propose a joint optimization method for post-processing the outputs of all modules using universal post-processing networks (UniPPNs), which are language-model-based networks that can modify the outputs of arbitrary modules in a system as a sequence-transformation task. Moreover, our RL algorithm, which employs a module-level Markov decision process, enables fine-grained value and advantage estimation for each module, thereby stabilizing joint learning for post-processing the outputs of all modules. Through both simulation-based and human evaluation experiments using the MultiWOZ dataset, we demonstrated that UniPPN outperforms conventional PPNs in the task completion capability of task-oriented dialogue systems.

Deep Reinforcement Learning (DRL) systems often tend to overfit to early experiences, a phenomenon known as the primacy bias (PB). This bias can severely hinder learning efficiency and final performance, particularly in complex environments. This paper presents a comprehensive investigation of PB through the lens of the Fisher Information Matrix (FIM). We develop a framework characterizing PB through distinct patterns in the FIM trace, identifying critical memorization and reorganization phases during learning. Building on this understanding, we propose Fisher-Guided Selective Forgetting (FGSF), a novel method that leverages the geometric structure of the parameter space to selectively modify network weights, preventing early experiences from dominating the learning process. Empirical results across DeepMind Control Suite (DMC) environments show that FGSF consistently outperforms baselines, particularly in complex tasks. We analyze the different impacts of PB on actor and critic networks, the role of replay ratios in exacerbating the effect, and the effectiveness of even simple noise injection methods. Our findings provide a deeper understanding of PB and practical mitigation strategies, offering a FIM-based geometric perspective for advancing DRL.

Monte Carlo Tree Search (MCTS) has proven highly effective in solving complex planning tasks by balancing exploration and exploitation using Upper Confidence Bound for Trees (UCT). However, existing work have not considered MCTS-based lifelong planning, where an agent faces a non-stationary series of tasks -- e.g., with varying transition probabilities and rewards -- that are drawn sequentially throughout the operational lifetime. This paper presents LiZero for Lipschitz lifelong planning using MCTS. We propose a novel concept of adaptive UCT (aUCT) to transfer knowledge from a source task to the exploration/exploitation of a new task, depending on both the Lipschitz continuity between tasks and the confidence of knowledge in in Monte Carlo action sampling. We analyze LiZero's acceleration factor in terms of improved sampling efficiency and also develop efficient algorithms to compute aUCT in an online fashion by both data-driven and model-based approaches, whose sampling complexity and error bounds are also characterized. Experiment results show that LiZero significantly outperforms existing MCTS and lifelong learning baselines in terms of much faster convergence (3$\sim$4x) to optimal rewards. Our results highlight the potential of LiZero to advance decision-making and planning in dynamic real-world environments.

Experience replay is a foundational technique in reinforcement learning that enhances learning stability by storing past experiences in a replay buffer and reusing them during training. Despite its practical success, its theoretical properties remain underexplored. In this paper, we present a theoretical framework that models experience replay using resampled $U$- and $V$-statistics, providing rigorous variance reduction guarantees. We apply this framework to policy evaluation tasks using the Least-Squares Temporal Difference (LSTD) algorithm and a Partial Differential Equation (PDE)-based model-free algorithm, demonstrating significant improvements in stability and efficiency, particularly in data-scarce scenarios. Beyond policy evaluation, we extend the framework to kernel ridge regression, showing that the experience replay-based method reduces the computational cost from the traditional $O(n^3)$ in time to as low as $O(n^2)$ in time while simultaneously reducing variance. Extensive numerical experiments validate our theoretical findings, demonstrating the broad applicability and effectiveness of experience replay in diverse machine learning tasks.

To bridge the gap between empirical success and theoretical understanding in transfer reinforcement learning (RL), we study a principled approach with provable performance guarantees. We introduce a novel composite MDP framework where high-dimensional transition dynamics are modeled as the sum of a low-rank component representing shared structure and a sparse component capturing task-specific variations. This relaxes the common assumption of purely low-rank transition models, allowing for more realistic scenarios where tasks share core dynamics but maintain individual variations. We introduce UCB-TQL (Upper Confidence Bound Transfer Q-Learning), designed for transfer RL scenarios where multiple tasks share core linear MDP dynamics but diverge along sparse dimensions. When applying UCB-TQL to a target task after training on a source task with sufficient trajectories, we achieve a regret bound of $\tilde{O}(\sqrt{eH^5N})$ that scales independently of the ambient dimension. Here, $N$ represents the number of trajectories in the target task, while $e$ quantifies the sparse differences between tasks. This result demonstrates substantial improvement over single task RL by effectively leveraging their structural similarities. Our theoretical analysis provides rigorous guarantees for how UCB-TQL simultaneously exploits shared dynamics while adapting to task-specific variations.

We study the benefits of different sparse architectures for deep reinforcement learning. In particular, we focus on image-based domains where spatially-biased and fully-connected architectures are common. Using these and several other architectures of equal capacity, we show that sparse structure has a significant effect on learning performance. We also observe that choosing the best sparse architecture for a given domain depends on whether the hidden layer weights are fixed or learned.

The significant role of division of labor (DOL) in promoting cooperation is widely recognized in real-world applications.Many cooperative multi-agent reinforcement learning (MARL) methods have incorporated the concept of DOL to improve cooperation among agents.However, the tasks used in existing testbeds typically correspond to tasks where DOL is often not a necessary feature for achieving optimal policies.Additionally, the full utilize of DOL concept in MARL methods remains unrealized due to the absence of appropriate tasks.To enhance the generality and applicability of MARL methods in real-world scenarios, there is a necessary to develop tasks that demand multi-agent DOL and cooperation.In this paper, we propose a series of tasks designed to meet these requirements, drawing on real-world rules as the guidance for their design.We guarantee that DOL and cooperation are necessary condition for completing tasks and introduce three factors to expand the diversity of proposed tasks to cover more realistic situations.We evaluate 10 cooperative MARL methods on the proposed tasks.The results indicate that all baselines perform poorly on these tasks.To further validate the solvability of these tasks, we also propose simplified variants of proposed tasks.Experimental results show that baselines are able to handle these simplified variants, providing evidence of the solvability of the proposed tasks.The source files is available at https://github.com/Yurui-Li/CTC.

The success of deep reinforcement learning (DRL) relies on the availability and quality of training data, often requiring extensive interactions with specific environments. In many real-world scenarios, where data collection is costly and risky, offline reinforcement learning (RL) offers a solution by utilizing data collected by domain experts and searching for a batch-constrained optimal policy. This approach is further augmented by incorporating external data sources, expanding the range and diversity of data collection possibilities. However, existing offline RL methods often struggle with challenges posed by non-matching data from these external sources. In this work, we specifically address the problem of source-target domain mismatch in scenarios involving mixed datasets, characterized by a predominance of source data generated from random or suboptimal policies and a limited amount of target data generated from higher-quality policies. To tackle this problem, we introduce Transition Scoring (TS), a novel method that assigns scores to transitions based on their similarity to the target domain, and propose Curriculum Learning-Based Trajectory Valuation (CLTV), which effectively leverages these transition scores to identify and prioritize high-quality trajectories through a curriculum learning approach. Our extensive experiments across various offline RL methods and MuJoCo environments, complemented by rigorous theoretical analysis, demonstrate that CLTV enhances the overall performance and transferability of policies learned by offline RL algorithms.