Markov decision processes (MDPs) are used to model a wide variety of applications ranging from game playing over robotics to finance. Their optimal policy typically maximizes the expected sum of rewards given at each step of the decision process. However, a large class of problems does not fit straightforwardly into this framework: Non-cumulative Markov decision processes (NCMDPs), where instead of the expected sum of rewards, the expected value of an arbitrary function of the rewards is maximized. Example functions include the maximum of the rewards or their mean divided by their standard deviation. In this work, we introduce a general mapping of NCMDPs to standard MDPs. This allows all techniques developed to find optimal policies for MDPs, such as reinforcement learning or dynamic programming, to be directly applied to the larger class of NCMDPs. Focusing on reinforcement learning, we show applications in a diverse set of tasks, including classical control, portfolio optimization in finance, and discrete optimization problems. Given our approach, we can improve both final performance and training time compared to relying on standard MDPs.
An informal observation, made by several authors, is that the adaptive design of a Markov transition kernel has the flavour of a reinforcement learning task. Yet, to-date it has remained unclear how to actually exploit modern reinforcement learning technologies for adaptive MCMC. The aim of this paper is to set out a general framework, called Reinforcement Learning Metropolis--Hastings, that is theoretically supported and empirically validated. Our principal focus is on learning fast-mixing Metropolis--Hastings transition kernels, which we cast as deterministic policies and optimise via a policy gradient. Control of the learning rate provably ensures conditions for ergodicity are satisfied. The methodology is used to construct a gradient-free sampler that out-performs a popular gradient-free adaptive Metropolis--Hastings algorithm on $\approx 90 \%$ of tasks in the PosteriorDB benchmark.
Reinforcement learning (RL) recommender systems often rely on static datasets that fail to capture the fluid, ever changing nature of user preferences in real-world scenarios. Meanwhile, generative AI techniques have emerged as powerful tools for creating synthetic data, including user profiles and behaviors. Recognizing this potential, we introduce Lusifer, an LLM-based simulation environment designed to generate dynamic, realistic user feedback for RL-based recommender training. In Lusifer, user profiles are incrementally updated at each interaction step, with Large Language Models (LLMs) providing transparent explanations of how and why preferences evolve. We focus on the MovieLens dataset, extracting only the last 40 interactions for each user, to emphasize recent behavior. By processing textual metadata (such as movie overviews and tags) Lusifer creates more context aware user states and simulates feedback on new items, including those with limited or no prior ratings. This approach reduces reliance on extensive historical data and facilitates cold start scenario handling and adaptation to out of distribution cases. Our experiments compare Lusifer with traditional collaborative filtering models, revealing that while Lusifer can be comparable in predictive accuracy, it excels at capturing dynamic user responses and yielding explainable results at every step. These qualities highlight its potential as a scalable, ethically sound alternative to live user experiments, supporting iterative and user-centric evaluations of RL-based recommender strategies. Looking ahead, we envision Lusifer serving as a foundational tool for exploring generative AI-driven user simulations, enabling more adaptive and personalized recommendation pipelines under real world constraints.
Designing generalizable agents capable of adapting to diverse embodiments has achieved significant attention in Reinforcement Learning (RL), which is critical for deploying RL agents in various real-world applications. Previous Cross-Embodiment RL approaches have focused on transferring knowledge across embodiments within specific tasks. These methods often result in knowledge tightly coupled with those tasks and fail to adequately capture the distinct characteristics of different embodiments. To address this limitation, we introduce the notion of Cross-Embodiment Unsupervised RL (CEURL), which leverages unsupervised learning to enable agents to acquire embodiment-aware and task-agnostic knowledge through online interactions within reward-free environments. We formulate CEURL as a novel Controlled Embodiment Markov Decision Process (CE-MDP) and systematically analyze CEURL's pre-training objectives under CE-MDP. Based on these analyses, we develop a novel algorithm Pre-trained Embodiment-Aware Control (PEAC) for handling CEURL, incorporating an intrinsic reward function specifically designed for cross-embodiment pre-training. PEAC not only provides an intuitive optimization strategy for cross-embodiment pre-training but also can integrate flexibly with existing unsupervised RL methods, facilitating cross-embodiment exploration and skill discovery. Extensive experiments in both simulated (e.g., DMC and Robosuite) and real-world environments (e.g., legged locomotion) demonstrate that PEAC significantly improves adaptation performance and cross-embodiment generalization, demonstrating its effectiveness in overcoming the unique challenges of CEURL. The project page and code are in https://yingchengyang.github.io/ceurl.
We introduce a novel reinforcement learning framework of LLM agents named AGILE (AGent that Interacts and Learns from Environments) designed to perform complex conversational tasks with users, leveraging LLMs, memory, tools, and interactions with experts. The agent possesses capabilities beyond conversation, including reflection, tool usage, and expert consultation. We formulate the construction of such an LLM agent as a reinforcement learning (RL) problem, in which the LLM serves as the policy model. We fine-tune the LLM using labeled data of actions and the PPO algorithm. We focus on question answering and release a dataset for agents called ProductQA, comprising challenging questions in online shopping. Our extensive experiments on ProductQA, MedMCQA and HotPotQA show that AGILE agents based on 7B and 13B LLMs trained with PPO can outperform GPT-4 agents. Our ablation study highlights the indispensability of memory, tools, consultation, reflection, and reinforcement learning in achieving the agent's strong performance. Datasets and code are available at https://github.com/bytarnish/AGILE.
Reinforcement Learning from Human Feedback (RLHF) has become the standard approach for aligning Large Language Models (LLMs) with human preferences, allowing LLMs to demonstrate remarkable abilities in various tasks. Existing methods work by emulating the preferences at the single decision (turn) level, limiting their capabilities in settings that require planning or multi-turn interactions to achieve a long-term goal. In this paper, we address this issue by developing novel methods for Reinforcement Learning (RL) from preference feedback between two full multi-turn conversations. In the tabular setting, we present a novel mirror-descent-based policy optimization algorithm for the general multi-turn preference-based RL problem, and prove its convergence to Nash equilibrium. To evaluate performance, we create a new environment, Education Dialogue, where a teacher agent guides a student in learning a random topic, and show that a deep RL variant of our algorithm outperforms RLHF baselines. Finally, we show that in an environment with explicit rewards, our algorithm recovers the same performance as a reward-based RL baseline, despite relying solely on a weaker preference signal.
Reinforcement learning (RL) is an effective method of finding reasoning pathways in incomplete knowledge graphs (KGs). To overcome the challenges of a large action space, a self-supervised pre-training method is proposed to warm up the policy network before the RL training stage. To alleviate the distributional mismatch issue in general self-supervised RL (SSRL), in our supervised learning (SL) stage, the agent selects actions based on the policy network and learns from generated labels; this self-generation of labels is the intuition behind the name self-supervised. With this training framework, the information density of our SL objective is increased and the agent is prevented from getting stuck with the early rewarded paths. Our self-supervised RL (SSRL) method improves the performance of RL by pairing it with the wide coverage achieved by SL during pretraining, since the breadth of the SL objective makes it infeasible to train an agent with that alone. We show that our SSRL model meets or exceeds current state-of-the-art results on all Hits@k and mean reciprocal rank (MRR) metrics on four large benchmark KG datasets. This SSRL method can be used as a plug-in for any RL architecture for a KGR task. We adopt two RL architectures, i.e., MINERVA and MultiHopKG as our baseline RL models and experimentally show that our SSRL model consistently outperforms both baselines on all of these four KG reasoning tasks. Full code for the paper available at https://github.com/owenonline/Knowledge-Graph-Reasoning-with-Self-supervised-Reinforcement-Learning.
Current Reinforcement Learning (RL) is often limited by the large amount of data needed to learn a successful policy. Offline RL aims to solve this issue by using transitions collected by a different behavior policy. We address a novel Offline RL problem setting in which, while collecting the dataset, the transition and reward functions gradually change between episodes but stay constant within each episode. We propose a method based on Contrastive Predictive Coding that identifies this non-stationarity in the offline dataset, accounts for it when training a policy, and predicts it during evaluation. We analyze our proposed method and show that it performs well in simple continuous control tasks and challenging, high-dimensional locomotion tasks. We show that our method often achieves the oracle performance and performs better than baselines.
Deep reinforcement learning (DRL) is a powerful machine learning paradigm for generating agents that control autonomous systems. However, the ``black box'' nature of DRL agents limits their deployment in real-world safety-critical applications. A promising approach for providing strong guarantees on an agent's behavior is to use Neural Lyapunov Barrier (NLB) certificates, which are learned functions over the system whose properties indirectly imply that an agent behaves as desired. However, NLB-based certificates are typically difficult to learn and even more difficult to verify, especially for complex systems. In this work, we present a novel method for training and verifying NLB-based certificates for discrete-time systems. Specifically, we introduce a technique for certificate composition, which simplifies the verification of highly-complex systems by strategically designing a sequence of certificates. When jointly verified with neural network verification engines, these certificates provide a formal guarantee that a DRL agent both achieves its goals and avoids unsafe behavior. Furthermore, we introduce a technique for certificate filtering, which significantly simplifies the process of producing formally verified certificates. We demonstrate the merits of our approach with a case study on providing safety and liveness guarantees for a DRL-controlled spacecraft.
World models constitute a promising approach for training reinforcement learning agents in a safe and sample-efficient manner. Recent world models predominantly operate on sequences of discrete latent variables to model environment dynamics. However, this compression into a compact discrete representation may ignore visual details that are important for reinforcement learning. Concurrently, diffusion models have become a dominant approach for image generation, challenging well-established methods modeling discrete latents. Motivated by this paradigm shift, we introduce DIAMOND (DIffusion As a Model Of eNvironment Dreams), a reinforcement learning agent trained in a diffusion world model. We analyze the key design choices that are required to make diffusion suitable for world modeling, and demonstrate how improved visual details can lead to improved agent performance. DIAMOND achieves a mean human normalized score of 1.46 on the competitive Atari 100k benchmark; a new best for agents trained entirely within a world model. We further demonstrate that DIAMOND's diffusion world model can stand alone as an interactive neural game engine by training on static Counter-Strike: Global Offensive gameplay. To foster future research on diffusion for world modeling, we release our code, agents, videos and playable world models at https://diamond-wm.github.io.
Transformer-based trajectory optimization methods have demonstrated exceptional performance in offline Reinforcement Learning (offline RL). Yet, it poses challenges due to substantial parameter size and limited scalability, which is particularly critical in sequential decision-making scenarios where resources are constrained such as in robots and drones with limited computational power. Mamba, a promising new linear-time sequence model, offers performance on par with transformers while delivering substantially fewer parameters on long sequences. As it remains unclear whether Mamba is compatible with trajectory optimization, this work aims to conduct comprehensive experiments to explore the potential of Decision Mamba (dubbed DeMa) in offline RL from the aspect of data structures and essential components with the following insights: (1) Long sequences impose a significant computational burden without contributing to performance improvements since DeMa's focus on sequences diminishes approximately exponentially. Consequently, we introduce a Transformer-like DeMa as opposed to an RNN-like DeMa. (2) For the components of DeMa, we identify the hidden attention mechanism as a critical factor in its success, which can also work well with other residual structures and does not require position embedding. Extensive evaluations demonstrate that our specially designed DeMa is compatible with trajectory optimization and surpasses previous methods, outperforming Decision Transformer (DT) with higher performance while using 30\% fewer parameters in Atari, and exceeding DT with only a quarter of the parameters in MuJoCo.
Reinforcement Learning (RL) algorithms often struggle with low training efficiency. A common approach to address this challenge is integrating model-based planning algorithms, such as Monte Carlo Tree Search (MCTS) or Value Iteration (VI), into the environmental model. However, VI requires iterating over a large tensor which updates the value of the preceding state based on the succeeding state through value propagation, resulting in computationally intensive operations. To enhance the RL training efficiency, we propose improving the efficiency of the value learning process. In deterministic environments with discrete state and action spaces, we observe that on the sampled empirical state-transition graph, a non-branching sequence of transitions-termed a highway-can take the agent to another state without deviation through intermediate states. On these non-branching highways, the value-updating process can be streamlined into a single-step operation, eliminating the need for step-by-step updates. Building on this observation, we introduce the highway graph to model state transitions. The highway graph compresses the transition model into a compact representation, where edges can encapsulate multiple state transitions, enabling value propagation across multiple time steps in a single iteration. By integrating the highway graph into RL, the training process is significantly accelerated, particularly in the early stages of training. Experiments across four categories of environments demonstrate that our method learns significantly faster than established and state-of-the-art RL algorithms (often by a factor of 10 to 150) while maintaining equal or superior expected returns. Furthermore, a deep neural network-based agent trained using the highway graph exhibits improved generalization capabilities and reduced storage costs. Code is publicly available at https://github.com/coodest/highwayRL.
In the field of safe reinforcement learning (RL), finding a balance between satisfying safety constraints and optimizing reward performance presents a significant challenge. A key obstacle in this endeavor is the estimation of safety constraints, which is typically more difficult than estimating a reward metric due to the sparse nature of the constraint signals. To address this issue, we introduce a novel framework named Feasibility Consistent Safe Reinforcement Learning (FCSRL). This framework combines representation learning with feasibility-oriented objectives to identify and extract safety-related information from the raw state for safe RL. Leveraging self-supervised learning techniques and a more learnable safety metric, our approach enhances the policy learning and constraint estimation. Empirical evaluations across a range of vector-state and image-based tasks demonstrate that our method is capable of learning a better safety-aware embedding and achieving superior performance than previous representation learning baselines.
Offline meta reinforcement learning (OMRL) has emerged as a promising approach for interaction avoidance and strong generalization performance by leveraging pre-collected data and meta-learning techniques. Previous context-based approaches predominantly rely on the intuition that alternating optimization between the context encoder and the policy can lead to performance improvements, as long as the context encoder follows the principle of maximizing the mutual information between the task variable $M$ and its latent representation $Z$ ($I(Z;M)$) while the policy adopts the standard offline reinforcement learning (RL) algorithms conditioning on the learned task representation.Despite promising results, the theoretical justification of performance improvements for such intuition remains underexplored.Inspired by the return discrepancy scheme in the model-based RL field, we find that the previous optimization framework can be linked with the general RL objective of maximizing the expected return, thereby explaining performance improvements. Furthermore, after scrutinizing this optimization framework, we observe that the condition for monotonic performance improvements does not consider the variation of the task representation. When these variations are considered, the previously established condition may no longer be sufficient to ensure monotonicity, thereby impairing the optimization process.We name this issue task representation shift and theoretically prove that the monotonic performance improvements can be guaranteed with appropriate context encoder updates.Our work opens up a new avenue for OMRL, leading to a better understanding between the task representation and performance improvements.
Multi-agent reinforcement learning (MARL) algorithms have accomplished remarkable breakthroughs in solving large-scale decision-making tasks. Nonetheless, most existing MARL algorithms are model-free, limiting sample efficiency and hindering their applicability in more challenging scenarios. In contrast, model-based reinforcement learning (MBRL), particularly algorithms integrating planning, such as MuZero, has demonstrated superhuman performance with limited data in many tasks. Hence, we aim to boost the sample efficiency of MARL by adopting model-based approaches. However, incorporating planning and search methods into multi-agent systems poses significant challenges. The expansive action space of multi-agent systems often necessitates leveraging the nearly-independent property of agents to accelerate learning. To tackle this issue, we propose the MAZero algorithm, which combines a centralized model with Monte Carlo Tree Search (MCTS) for policy search. We design a novel network structure to facilitate distributed execution and parameter sharing. To enhance search efficiency in deterministic environments with sizable action spaces, we introduce two novel techniques: Optimistic Search Lambda (OS($\lambda$)) and Advantage-Weighted Policy Optimization (AWPO). Extensive experiments on the SMAC benchmark demonstrate that MAZero outperforms model-free approaches in terms of sample efficiency and provides comparable or better performance than existing model-based methods in terms of both sample and computational efficiency. Our code is available at https://github.com/liuqh16/MAZero.
Over the last decade, data-driven methods have surged in popularity, emerging as valuable tools for control theory. As such, neural network approximations of control feedback laws, system dynamics, and even Lyapunov functions have attracted growing attention. With the ascent of learning based control, the need for accurate, fast, and easy-to-use benchmarks has increased. In this work, we present the first learning-based environment for boundary control of PDEs. In our benchmark, we introduce three foundational PDE problems - a 1D transport PDE, a 1D reaction-diffusion PDE, and a 2D Navier-Stokes PDE - whose solvers are bundled in an user-friendly reinforcement learning gym. With this gym, we then present the first set of model-free, reinforcement learning algorithms for solving this series of benchmark problems, achieving stability, although at a higher cost compared to model-based PDE backstepping. With the set of benchmark environments and detailed examples, this work significantly lowers the barrier to entry for learning-based PDE control - a topic largely unexplored by the data-driven control community. The entire benchmark is available on Github along with detailed documentation and the presented reinforcement learning models are open sourced.
This paper presents a study of robust policy networks in deep reinforcement learning. We investigate the benefits of policy parameterizations that naturally satisfy constraints on their Lipschitz bound, analyzing their empirical performance and robustness on two representative problems: pendulum swing-up and Atari Pong. We illustrate that policy networks with smaller Lipschitz bounds are more robust to disturbances, random noise, and targeted adversarial attacks than unconstrained policies composed of vanilla multi-layer perceptrons or convolutional neural networks. However, the structure of the Lipschitz layer is important. We find that the widely-used method of spectral normalization is too conservative and severely impacts clean performance, whereas more expressive Lipschitz layers such as the recently-proposed Sandwich layer can achieve improved robustness without sacrificing clean performance.
As large language models (LLMs) continue to grow by scaling laws, reinforcement learning from human feedback (RLHF) has gained significant attention due to its outstanding performance. However, unlike pretraining or fine-tuning a single model, scaling reinforcement learning from human feedback (RLHF) for training large language models poses coordination challenges across four models. We present OpenRLHF, an open-source framework enabling efficient RLHF scaling. Unlike existing RLHF frameworks that co-locate four models on the same GPUs, OpenRLHF re-designs scheduling for the models beyond 70B parameters using Ray, vLLM, and DeepSpeed, leveraging improved resource utilization and diverse training approaches. Integrating seamlessly with Hugging Face, OpenRLHF provides an out-of-the-box solution with optimized algorithms and launch scripts, which ensures user-friendliness. OpenRLHF implements RLHF, DPO, rejection sampling, and other alignment techniques. Empowering state-of-the-art LLM development, OpenRLHF's code is available at \url{https://github.com/OpenRLHF/OpenRLHF}.
Observing celestial objects and advancing our scientific knowledge about them involves tedious planning, scheduling, data collection and data post-processing. Many of these operational aspects of astronomy are guided and executed by expert astronomers. Reinforcement learning is a mechanism where we (as humans and astronomers) can teach agents of artificial intelligence to perform some of these tedious tasks. In this paper, we will present a state of the art overview of reinforcement learning and how it can benefit astronomy.
We show that discounted methods for solving continuing reinforcement learning problems can perform significantly better if they center their rewards by subtracting out the rewards' empirical average. The improvement is substantial at commonly used discount factors and increases further as the discount factor approaches one. In addition, we show that if a problem's rewards are shifted by a constant, then standard methods perform much worse, whereas methods with reward centering are unaffected. Estimating the average reward is straightforward in the on-policy setting; we propose a slightly more sophisticated method for the off-policy setting. Reward centering is a general idea, so we expect almost every reinforcement-learning algorithm to benefit by the addition of reward centering.