Large transformer models, trained on diverse datasets, have demonstrated impressive few-shot performance on previously unseen tasks without requiring parameter updates. This capability has also been explored in Reinforcement Learning (RL), where agents interact with the environment to retrieve context and maximize cumulative rewards, showcasing strong adaptability in complex settings. However, in cooperative Multi-Agent Reinforcement Learning (MARL), where agents must coordinate toward a shared goal, decentralized policy deployment can lead to mismatches in task alignment and reward assignment, limiting the efficiency of policy adaptation. To address this challenge, we introduce Multi-agent In-context Coordination via Decentralized Memory Retrieval (MAICC), a novel approach designed to enhance coordination by fast adaptation. Our method involves training a centralized embedding model to capture fine-grained trajectory representations, followed by decentralized models that approximate the centralized one to obtain team-level task information. Based on the learned embeddings, relevant trajectories are retrieved as context, which, combined with the agents' current sub-trajectories, inform decision-making. During decentralized execution, we introduce a novel memory mechanism that effectively balances test-time online data with offline memory. Based on the constructed memory, we propose a hybrid utility score that incorporates both individual- and team-level returns, ensuring credit assignment across agents. Extensive experiments on cooperative MARL benchmarks, including Level-Based Foraging (LBF) and SMAC (v1/v2), show that MAICC enables faster adaptation to unseen tasks compared to existing methods.

Large Language Models (LLMs), trained on a large amount of corpus, have demonstrated remarkable abilities. However, it may not be sufficient to directly apply open-source LLMs like Llama to certain real-world scenarios, since most of them are trained for \emph{general} purposes. Thus, the demands for customizing publicly available LLMs emerge, but are currently under-studied. In this work, we consider customizing pre-trained LLMs with new human preferences. Specifically, the LLM should not only meet the new preference but also preserve its original capabilities after customization. Drawing inspiration from the observation that human preference can be expressed as a reward model, we propose to cast LLM customization as optimizing the sum of two reward functions, one of which (denoted as $r_1$) was used to pre-train the LLM while the other (denoted as $r_2$) characterizes the new human preference. The obstacle here is that both reward functions are unknown, making the application of modern reinforcement learning methods infeasible. Thanks to the residual Q-learning framework, we can restore the customized LLM with the pre-trained LLM and the \emph{residual Q-function} without the reward function $r_1$. Moreover, we find that for a fixed pre-trained LLM, the reward function $r_2$ can be derived from the residual Q-function, enabling us to directly learn the residual Q-function from the new human preference data upon the Bradley-Terry model. We name our method Q-Adapter as it introduces an adapter module to approximate the residual Q-function for customizing the pre-trained LLM towards the new preference. Experiments based on the Llama-3.1 model on the DSP dataset and HH-RLHF dataset illustrate the superior effectiveness of Q-Adapter on both retaining existing knowledge and learning new preferences. Our code is available at \url{https://github.com/LAMDA-RL/Q-Adapter}.

Model-based methods in reinforcement learning offer a promising approach to enhance data efficiency by facilitating policy exploration within a dynamics model. However, accurately predicting sequential steps in the dynamics model remains a challenge due to the bootstrapping prediction, which attributes the next state to the prediction of the current state. This leads to accumulated errors during model roll-out. In this paper, we propose the Any-step Dynamics Model (ADM) to mitigate the compounding error by reducing bootstrapping prediction to direct prediction. ADM allows for the use of variable-length plans as inputs for predicting future states without frequent bootstrapping. We design two algorithms, ADMPO-ON and ADMPO-OFF, which apply ADM in online and offline model-based frameworks, respectively. In the online setting, ADMPO-ON demonstrates improved sample efficiency compared to previous state-of-the-art methods. In the offline setting, ADMPO-OFF not only demonstrates superior performance compared to recent state-of-the-art offline approaches but also offers better quantification of model uncertainty using only a single ADM.

Adapting Large Language Models (LLMs) to downstream tasks using Reinforcement Learning (RL) has proven to be an effective approach. However, LLMs do not inherently define the structure of an agent for RL training, particularly in terms of specifying the action space. This paper studies learning a compact latent action space to enhance the controllability and exploration of RL for LLMs. Inspired by reinforcement learning from observations, we propose Controlling Large Language Models with Latent Actions CoLA, a framework that integrates a latent action space into pre-trained LLMs. CoLA employs an inverse dynamics model to extract latent actions conditioned on future tokens, ensuring that the next token prediction is partially influenced by these actions. Simultaneously, CoLA fine-tunes the pre-trained LLM to function as a language world model, capable of incorporating latent actions as inputs. Additionally, CoLA trains a policy model to generate actions within this language world model. The policy model can be trained via behavior cloning to mimic a standard language model or through RL to maximize task-specific rewards. In this work, we apply CoLA to the Llama-3.1-8B model. Our experiments demonstrate that, compared to RL with token-level actions, CoLA’s latent actions enable greater semantic diversity. For enhancing downstream tasks, we show that CoLA with RL achieves a score of 42.4 on the math500 benchmark, surpassing the baseline score of 38.2, and reaches 68.2 when augmented with a Monte Carlo Tree Search variant. Furthermore, CoLA with RL consistently improves performance on agent-based tasks without degrading the pre-trained LLM’s capabilities, unlike the baseline. Finally, CoLA reduces computation time by half in tasks involving enhanced thinking prompts for LLMs via RL. These results highlight CoLA’s potential to advance RL-based adaptation of LLMs for downstream applications. The CoLA model is available at https://huggingface.co/LAMDA-RL/Llama-3.1-CoLA-10B.

The policy trained via reinforcement learning (RL) makes decisions based on sensor-derived state features. It is common for state features to evolve for reasons such as periodic sensor maintenance or the addition of new sensors for performance improvement. The deployed policy fails in new state space when state features are unseen during training. Previous work tackles this challenge by training a sensor-invariant policy or generating multiple policies and selecting the appropriate one with limited samples. However, both directions struggle to guarantee the performance when faced with unpredictable evolutions. In this paper, we formalize this problem as state evolvable reinforcement learning (SERL), where the agent is required to mitigate policy degradation after state evolutions without costly exploration. We propose Lapse by reusing policies learned from the old state space in two distinct aspects. On one hand, Lapse directly reuses the robust old policy by composing it with a learned state reconstruction model to handle vanishing sensors. On the other hand, the behavioral experience from the old policy is reused by Lapse to train a newly adaptive policy through offline learning, better utilizing new sensors. To leverage advantages of both policies in different scenarios, we further propose automatic ensemble weight adjustment to effectively aggregate them. Theoretically, we justify that robust policy reuse helps mitigate uncertainty and error from both evolution and reconstruction. Empirically, Lapse achieves a significant performance improvement, outperforming the strongest baseline by about $2\times$ in benchmark environments.

Learning from multi-agent expert demonstrations, known as Multi-Agent Imitation Learning (MAIL), provides a promising approach to sequential decision-making. However, existing MAIL methods including Behavior Cloning (BC) and Adversarial Imitation Learning (AIL) face significant challenges: BC suffers from the compounding error issue, while the very nature of adversarial optimization makes AIL prone to instability. In this work, we propose \textbf{M}ulti-\textbf{A}gent imitation by learning and sampling from \textbf{F}actor\textbf{I}zed \textbf{S}oft Q-function (MAFIS), a novel method that addresses these limitations for both online and offline MAIL settings. Built upon the single-agent IQ-Learn framework, MAFIS introduces the value decomposition network to factorize the imitation objective at agent level, thus enabling scalable training for multi-agent systems. Moreover, we observe that the soft Q-function implicitly defines the optimal policy as an energy-based model, from which we can sample actions via stochastic gradient Langevin dynamics. This allows us to estimate the gradient of the factorized optimization objective for continuous control tasks, avoiding the adversarial optimization between the soft Q-function and the policy required by prior work. By doing so, we obtain a tractable and \emph{non-adversarial} objective for both discrete and continuous multi-agent control. Experiments on common benchmarks including the discrete control tasks StarCraft Multi-Agent Challenge v2 (SMACv2), Gold Miner, and Multi Particle Environments (MPE), as well as the continuous control task Multi-Agent MuJoCo (MaMuJoCo), demonstrate that MAFIS achieves superior performance compared with baselines. Our code is available at https://github.com/LAMDA-RL/MAFIS.

Imitation Learning (IL) learns from experts, on which most existing studies assume that the imitator will be deployed in stationary environments. However, real-world scenarios commonly involve perturbations, necessitating robust imitators for non-stationary scenarios. To fulfill this, we leverage a multi-modal expert dataset encompassing diverse dynamics, while still adhering to the shared goal between the experts and imitator. Different from conventional multi-modal IL work that considers reproducing the demonstrated different behaviors, we aim to imitate a policy that rapidly adapts to sudden dynamic changes, even when encountering dynamics unseen during training. We propose a method called Generalizable Multi-modal Adversarial Imitation Learning (GMAIL) for non-stationary dynamics, which adversarially trains a discriminator and a generator. Due to dynamic mismatch between the experts and the imitator, the optimal next state for the imitator may require several steps for the experts to reach, inspiring us to propose to take the state-next-state pairs within multiple steps in the demonstrated trajectories to facilitate imitation under dynamic mismatch. For quick identification of the changed dynamic, GMAIL learns a dynamics-sensitive generator by introducing a history-based context encoder. On a wide range of navigation, locomotion and autonomous driving tasks, empirical results illustrate the effectiveness of GMAIL.

Multiobjective reinforcement learning (MORL) addresses sequential decision-making problems with multiple objectives by learning policies optimized for diverse pReferences. While traditional methods necessitate costly online interaction with the environment, recent approaches leverage static datasets containing precollected trajectories, making offline MORL the preferred choice for real-world applications. However, existing offline MORL techniques suffer from limited expressiveness and poor generalization on out-of-distribution (OOD) preferences. To overcome these limitations, we propose diffusion-based MORL (DiffMORL), a generalizable diffusion-based planning frame work for MORL. Leveraging the strong expressiveness and generation capability of diffusion models, DiffMORL further boosts its generalization through offline data mixup, which mitigates the memorization phenomenon and facilitates feature learning by data augmentation. By training on the augmented data, DiffMORL is able to condition on a given preference, whether in-distribution or OOD, to plan the desired trajectory and extract the corresponding action. Evaluations conducted on the datasets for MORL (D4MORL) benchmark demonstrate that DiffMORL achieves state-of-the-art results across nearly all tasks. Notably, it surpasses the best baseline on 14 out of 18 metrics for OOD generalization, underscoring its remarkable generalization ability in offline MORL scenarios.

Low sample efficiency is an enduring challenge of reinforcement learning (RL). With the advent of versatile large language models (LLMs), recent works impart common-sense knowledge to accelerate policy learning for RL processes. However, we note that such guidance is often tailored for one specific task but loses generalizability. In this article, we introduce a framework that harnesses LLMs to extract background knowledge of an environment, which contains general understandings of the entire environment, making various downstream RL tasks benefit from one-time knowledge representation. We ground LLMs by feeding a few precollected experiences and requesting them to delineate background knowledge of the environment. Afterward, we represent the output knowledge as potential functions for potential-based reward shaping, which has a good property for maintaining policy optimality from task rewards. We instantiate three variants to prompt LLMs for background knowledge, including writing code, annotating pReferences, and assigning goals. Our experiments show that these methods achieve significant sample efficiency improvements in a spectrum of downstream tasks from Minigrid and Crafter domains.

One-shot imitation learning (OSIL) is to learn an imitator agent that can execute multiple tasks with only a single demonstration. In real-world scenario, the environment is dynamic, e.g., unexpected changes can occur after demonstration. Thus, achieving generalization of the imitator agent is crucial as agents would inevitably face situations unseen in the provided demonstrations. While traditional OSIL methods excel in relatively stationary settings, their adaptability to such unforeseen changes, which asking for a higher level of generalization ability for the imitator agents, is limited and rarely discussed. In this work, we present a new algorithm called Deep Demonstration Tracing (DDT). In DDT, we propose a demonstration transformer architecture to encourage agents to adaptively trace suitable states in demonstrations. Besides, it integrates OSIL into a meta-reinforcement-learning training paradigm, providing regularization for policies in unexpected situations. We evaluate DDT on a new navigation task suite and robotics tasks, demonstrating its superior performance over existing OSIL methods across all evaluated tasks in dynamic environments with unforeseen changes. The project page is in https://osil-ddt.github.io.

Developing policies that can adapt to non-stationary environments is essential for real-world reinforcement learning applications. Nevertheless, learning such adaptable policies in offline settings, with only a limited set of pre-collected trajectories, presents significant challenges. A key difficulty arises because the limited offline data makes it hard for the context encoder to differentiate between changes in the environment dynamics and shifts in the behavior policy, often leading to context misassociations. To address this issue, we introduce a novel approach called debiased offline representation learning for fast online adaptation (DORA). DORA incorporates an information bottleneck principle that maximizes mutual information between the dynamics encoding and the environmental data, while minimizing mutual information between the dynamics encoding and the actions of the behavior policy. We present a practical implementation of DORA, leveraging tractable bounds of the information bottleneck principle. Our experimental evaluation across six benchmark MuJoCo tasks with variable parameters demonstrates that DORA not only achieves a more precise dynamics encoding but also significantly outperforms existing baselines in terms of performance.

To handle a large amount of unlabeled data, batch active learning (BAL) queries humans for the labels of a batch of the most valuable data points at every round. Most current BAL strategies are based on human-designed heuristics, such as uncertainty sampling or mutual information maximization. However, there exists a disagreement between these heuristics and the ultimate goal of BAL, i.e., optimizing the model's final performance within the query budgets. This disagreement leads to a limited generality of these heuristics. To this end, we formulate BAL as an MDP and propose a data-driven approach based on deep reinforcement learning. Our method learns the BAL strategy by maximizing the model's final performance. Experiments on the UCI benchmark show that our method can achieve competitive performance compared to existing heuristics-based approaches.

We consider the problem of learning the best possible policy from a fixed dataset, known as offline Reinforcement Learning (RL). A common taxonomy of existing offline RL works is policy regularization, which typically constrains the learned policy by distribution or support of the behavior policy. However, distribution and support constraints are overly conservative since they both force the policy to choose similar actions as the behavior policy when considering particular states. It will limit the learned policy’s performance, especially when the behavior policy is sub-optimal. In this paper, we find that regularizing the policy towards the nearest state-action pair can be more effective and thus propose Policy Regularization with Dataset Constraint (PRDC). When updating the policy in a given state, PRDC searches the entire dataset for the nearest state-action sample and then restricts the policy with the action of this sample. Unlike previous works, PRDC can guide the policy with proper behaviors from the dataset, allowing it to choose actions that do not appear in the dataset along with the given state. It is a softer constraint but still keeps enough conservatism from out-of-distribution actions. Empirical evidence and theoretical analysis show that PRDC can alleviate offline RL’s fundamentally challenging value overestimation issue with a bounded performance gap. Moreover, on a set of locomotion and navigation tasks, PRDC achieves state-of-the-art performance compared with existing methods. Code is available at https://github.com/LAMDA-RL/PRDC