Multi-agent reinforcement learning (MARL) faces challenges in coordinating agents due to complex interdependencies within multi-agent systems. Most MARL algorithms use the simultaneous decision-making paradigm but ignore the action-level dependencies among agents, which reduces coordination efficiency. In contrast, the sequential decision-making paradigm provides finer-grained supervision for agent decision order, presenting the potential for handling dependencies via better decision order management. However, determining the optimal decision order remains a challenge. In this paper, we introduce Action Generation with Plackett-Luce Sampling (AGPS), a novel mechanism for agent decision order optimization. We model the order determination task as a Plackett-Luce sampling process to address issues such as ranking instability and vanishing gradient during the network training process. AGPS realizes credit-based decision order determination by establishing a bridge between the significance of agents' local observations and their decision credits, thus facilitating order optimization and dependency management. Integrating AGPS with the Multi-Agent Transformer, we propose the Prioritized Multi-Agent Transformer (PMAT), a sequential decision-making MARL algorithm with decision order optimization. Experiments on benchmarks including StarCraft II Multi-Agent Challenge, Google Research Football, and Multi-Agent MuJoCo show that PMAT outperforms state-of-the-art algorithms, greatly enhancing coordination efficiency.
Over the last decade, the use of machine learning (ML) approaches in medicinal applications has increased manifold. Most of these approaches are based on deep learning, which aims to learn representations from grid data (like medical images). However, reinforcement learning (RL) applications in medicine are relatively less explored. Medical applications often involve a sequence of subtasks that form a diagnostic pipeline, and RL is uniquely suited to optimize over such sequential decision-making tasks. Ultrasound (US) image analysis is a quintessential example of such a sequential decision-making task, where the raw signal captured by the US transducer undergoes a series of signal processing and image post-processing steps, generally leading to a diagnostic suggestion. The application of RL in US remains limited. Deep Reinforcement Learning (DRL), that combines deep learning and RL, holds great promise in optimizing these pipelines by enabling intelligent and sequential decision-making. This review paper surveys the applications of RL in US over the last decade. We provide a succinct overview of the theoretic framework of RL and its application in US image processing and review existing work in each aspect of the image analysis pipeline. A comprehensive search of Scopus filtered on relevance yielded 14 papers most relevant to this topic. These papers were further categorized based on their target applications image classification, image segmentation, image enhancement, video summarization, and auto navigation and path planning. We also examined the type of RL approach used in each publication. Finally, we discuss key areas in healthcare where DRL approaches in US could be used for sequential decision-making. We analyze the opportunities, challenges, and limitations, providing insights into the future potential of DRL in US image analysis.
The development of a generalist agent capable of solving a wide range of sequential decision-making tasks remains a significant challenge. We address this problem in a cross-agent setup where agents share the same observation space but differ in their action spaces. Our approach builds on the universal policy framework, which decouples policy learning into two stages: a diffusion-based planner that generates observation sequences and an inverse dynamics model that assigns actions to these plans. We propose a method for training the planner on a joint dataset composed of trajectories from all agents. This method offers the benefit of positive transfer by pooling data from different agents, while the primary challenge lies in adapting shared plans to each agent's unique constraints. We evaluate our approach on the BabyAI environment, covering tasks of varying complexity, and demonstrate positive transfer across agents. Additionally, we examine the planner's generalisation ability to unseen agents and compare our method to traditional imitation learning approaches. By training on a pooled dataset from multiple agents, our universal policy achieves an improvement of up to $42.20\%$ in task completion accuracy compared to a policy trained on a dataset from a single agent.
Large Language Models (LLMs) have demonstrated impressive capabilities in language processing, yet they often struggle with tasks requiring genuine visual spatial reasoning. In this paper, we introduce a novel two-stage training framework designed to equip standard LLMs with visual reasoning abilities for maze navigation. First, we leverage Supervised Fine Tuning (SFT) on a curated dataset of tokenized maze representations to teach the model to predict step-by-step movement commands. Next, we apply Group Relative Policy Optimization (GRPO)-a technique used in DeepSeekR1-with a carefully crafted reward function to refine the model's sequential decision-making and encourage emergent chain-of-thought behaviors. Experimental results on synthetically generated mazes show that while a baseline model fails to navigate the maze, the SFT-trained model achieves 86% accuracy, and further GRPO fine-tuning boosts accuracy to 93%. Qualitative analyses reveal that GRPO fosters more robust and self-corrective reasoning, highlighting the potential of our approach to bridge the gap between language models and visual spatial tasks. These findings offer promising implications for applications in robotics, autonomous navigation, and other domains that require integrated visual and sequential reasoning.
We propose the Value Gradient Sampler (VGS), a trainable sampler based on the interpretation of sampling as discrete-time sequential decision-making. VGS generates samples from a given unnormalized density (i.e., energy) by drifting and diffusing randomly initialized particles. In VGS, finding the optimal drift is equivalent to solving an optimal control problem where the cost is the upper bound of the KL divergence between the target density and the samples. We employ value-based dynamic programming to solve this optimal control problem, which gives the gradient of the value function as the optimal drift vector. The connection to sequential decision making allows VGS to leverage extensively studied techniques in reinforcement learning, making VGS a fast, adaptive, and accurate sampler that achieves competitive results in various sampling benchmarks. Furthermore, VGS can replace MCMC in contrastive divergence training of energy-based models. We demonstrate the effectiveness of VGS in training accurate energy-based models in industrial anomaly detection applications.
Infectious diseases pose major public health challenges to society, highlighting the importance of designing effective policies to reduce economic loss and mortality. In this paper, we propose a framework for sequential decision-making under uncertainty to design fairness-aware disease mitigation policies that incorporate various measures of unfairness. Specifically, our approach learns equitable vaccination and lockdown strategies based on a stochastic multi-group SIR model. To address the challenges of solving the resulting sequential decision-making problem, we adopt the path integral control algorithm as an efficient solution scheme. Through a case study, we demonstrate that our approach effectively improves fairness compared to conventional methods and provides valuable insights for policymakers.
The Mixed-Shelves Picker Routing Problem (MSPRP) is a fundamental challenge in warehouse logistics, where pickers must navigate a mixed-shelves environment to retrieve SKUs efficiently. Traditional heuristics and optimization-based approaches struggle with scalability, while recent machine learning methods often rely on sequential decision-making, leading to high solution latency and suboptimal agent coordination. In this work, we propose a novel hierarchical and parallel decoding approach for solving the min-max variant of the MSPRP via multi-agent reinforcement learning. While our approach generates a joint distribution over agent actions, allowing for fast decoding and effective picker coordination, our method introduces a sequential action selection to avoid conflicts in the multi-dimensional action space. Experiments show state-of-the-art performance in both solution quality and inference speed, particularly for large-scale and out-of-distribution instances. Our code is publicly available at http://github.com/LTluttmann/marl4msprp.
State-of-the-art (SOTA) reinforcement learning (RL) methods have enabled vision-language model (VLM) agents to learn from interaction with online environments without human supervision. However, these methods often struggle with learning inefficiencies when applied to complex, real-world decision-making tasks with sparse rewards and long-horizon dependencies. We propose a novel framework, Variational Subgoal-Conditioned Reinforcement Learning (VSC-RL), advancing the VLM agents in resolving challenging decision-making tasks. Fundamentally distinct from existing methods, VSC-RL reformulates the decision-making problem as a variational subgoal-conditioned RL problem with the newly derived optimization objective, Subgoal Evidence Lower BOund (SGC-ELBO), which comprises two key components: (a) maximizing the subgoal-conditioned return, and (b) minimizing the divergence from a reference goal-conditioned policy. We theoretically and empirically demonstrate that the VSC-RL can efficiently improve the learning efficiency without compromising performance guarantees. Across a diverse set of challenging benchmarks, including mobile device and web control tasks, VSC-RL consistently outperforms existing SOTA methods, achieving superior learning efficiency and performance.
Combinatorial optimization problems, such as scheduling and route planning, are crucial in various industries but are computationally intractable due to their NP-hard nature. Neural Combinatorial Optimization methods leverage machine learning to address these challenges but often depend on sequential decision-making, which is prone to error accumulation as small mistakes propagate throughout the process. Inspired by self-evaluation techniques in Large Language Models, we propose a novel framework that generates and evaluates subsets of assignments, moving beyond traditional stepwise approaches. Applied to the Job-Shop Scheduling Problem, our method integrates a heterogeneous graph neural network with a Transformer to build a policy model and a self-evaluation function. Experimental validation on challenging, well-known benchmarks demonstrates the effectiveness of our approach, surpassing state-of-the-art methods.
Deep Reinforcement Learning (DRL) has achieved remarkable success in sequential decision-making tasks across diverse domains, yet its reliance on black-box neural architectures hinders interpretability, trust, and deployment in high-stakes applications. Explainable Deep Reinforcement Learning (XRL) addresses these challenges by enhancing transparency through feature-level, state-level, dataset-level, and model-level explanation techniques. This survey provides a comprehensive review of XRL methods, evaluates their qualitative and quantitative assessment frameworks, and explores their role in policy refinement, adversarial robustness, and security. Additionally, we examine the integration of reinforcement learning with Large Language Models (LLMs), particularly through Reinforcement Learning from Human Feedback (RLHF), which optimizes AI alignment with human preferences. We conclude by highlighting open research challenges and future directions to advance the development of interpretable, reliable, and accountable DRL systems.
In online advertising, advertisers commonly utilize auto-bidding services to bid for impression opportunities. A typical objective of the auto-bidder is to optimize the advertiser's cumulative value of winning impressions within specified budget constraints. However, such a problem is challenging due to the complex bidding environment faced by diverse advertisers. To address this challenge, we introduce ABPlanner, a few-shot adaptable budget planner designed to improve budget-constrained auto-bidding. ABPlanner is based on a hierarchical bidding framework that decomposes the bidding process into shorter, manageable stages. Within this framework, ABPlanner allocates the budget across all stages, allowing a low-level auto-bidder to bids based on the budget allocation plan. The adaptability of ABPlanner is achieved through a sequential decision-making approach, inspired by in-context reinforcement learning. For each advertiser, ABPlanner adjusts the budget allocation plan episode by episode, using data from previous episodes as prompt for current decisions. This enables ABPlanner to quickly adapt to different advertisers with few-shot data, providing a sample-efficient solution. Extensive simulation experiments and real-world A/B testing validate the effectiveness of ABPlanner, demonstrating its capability to enhance the cumulative value achieved by auto-bidders.
Data selection has emerged as a crucial downstream application of data valuation. While existing data valuation methods have shown promise in selection tasks, the theoretical foundations and full potential of using data values for selection remain largely unexplored. In this work, we first demonstrate that data values applied for selection can be naturally reformulated as a sequential-decision-making problem, where the optimal data value can be derived through dynamic programming. We show this framework unifies and reinterprets existing methods like Data Shapley through the lens of approximate dynamic programming, specifically as myopic reward function approximations to this sequential problem. Furthermore, we analyze how sequential data selection optimality is affected when the ground-truth utility function exhibits monotonic submodularity with curvature. To address the computational challenges in obtaining optimal data values, we propose an efficient approximation scheme using learned bipartite graphs as surrogate utility models, ensuring greedy selection is still optimal when the surrogate utility is correctly specified and learned. Extensive experiments demonstrate the effectiveness of our approach across diverse datasets.
The contextual multi-armed bandit (MAB) is a widely used framework for problems requiring sequential decision-making under uncertainty, such as recommendation systems. In applications involving a large number of users, the performance of contextual MAB can be significantly improved by facilitating collaboration among multiple users. This has been achieved by the clustering of bandits (CB) methods, which adaptively group the users into different clusters and achieve collaboration by allowing the users in the same cluster to share data. However, classical CB algorithms typically rely on numerical reward feedback, which may not be practical in certain real-world applications. For instance, in recommendation systems, it is more realistic and reliable to solicit preference feedback between pairs of recommended items rather than absolute rewards. To address this limitation, we introduce the first "clustering of dueling bandit algorithms" to enable collaborative decision-making based on preference feedback. We propose two novel algorithms: (1) Clustering of Linear Dueling Bandits (COLDB) which models the user reward functions as linear functions of the context vectors, and (2) Clustering of Neural Dueling Bandits (CONDB) which uses a neural network to model complex, non-linear user reward functions. Both algorithms are supported by rigorous theoretical analyses, demonstrating that user collaboration leads to improved regret bounds. Extensive empirical evaluations on synthetic and real-world datasets further validate the effectiveness of our methods, establishing their potential in real-world applications involving multiple users with preference-based feedback.
Vision-Language-Action (VLA) model can process instructions and visual perception to directly generate actions as output in an end-to-end fashion due to its strong multi-modal reasoning capabilities. While the performance of VLA models is promising, their computational cost can be substantial. This raises challenge for applying them on robotics tasks, which requires real-time decision-making to respond quickly to environmental changes. Since robotic control involves sequential decision-making, the visual input often exhibits minimal variation between successive steps. A natural idea is to reuse the computational results of unchanged visual tokens from the last step. Motivated by this idea, we propose VLA-Cache, an efficient vision-language-action model. VLA-Cache incorporates a token-selection mechanism that compares the visual input at each step with the input from the previous step, adaptively identifying visual tokens with minimal changes. The computational results for these unchanged tokens are then reused in subsequent steps via KV-cache, thereby significantly improving the efficiency of the VLA-Cache model. Experimental results on both simulation (e.g., LIBERO benchmark and SIMPLER) and real-world robot valid VLA-Cache can achieve practical acceleration with minimal sacrifice in success rate.
Since the explosion in popularity of ChatGPT, large language models (LLMs) have continued to impact our everyday lives. Equipped with external tools that are designed for a specific purpose (e.g., for flight booking or an alarm clock), LLM agents exercise an increasing capability to assist humans in their daily work. Although LLM agents have shown a promising blueprint as daily assistants, there is a limited understanding of how they can provide daily assistance based on planning and sequential decision making capabilities. We draw inspiration from recent work that has highlighted the value of 'LLM-modulo' setups in conjunction with humans-in-the-loop for planning tasks. We conducted an empirical study (N = 248) of LLM agents as daily assistants in six commonly occurring tasks with different levels of risk typically associated with them (e.g., flight ticket booking and credit card payments). To ensure user agency and control over the LLM agent, we adopted LLM agents in a plan-then-execute manner, wherein the agents conducted step-wise planning and step-by-step execution in a simulation environment. We analyzed how user involvement at each stage affects their trust and collaborative team performance. Our findings demonstrate that LLM agents can be a double-edged sword -- (1) they can work well when a high-quality plan and necessary user involvement in execution are available, and (2) users can easily mistrust the LLM agents with plans that seem plausible. We synthesized key insights for using LLM agents as daily assistants to calibrate user trust and achieve better overall task outcomes. Our work has important implications for the future design of daily assistants and human-AI collaboration with LLM agents.
Autonomous systems are increasingly expected to operate in the presence of adversaries, though an adversary may infer sensitive information simply by observing a system, without even needing to interact with it. Therefore, in this work we present a deceptive decision-making framework that not only conceals sensitive information, but in fact actively misleads adversaries about it. We model autonomous systems as Markov decision processes, and we consider adversaries that attempt to infer their reward functions using inverse reinforcement learning. To counter such efforts, we present two regularization strategies for policy synthesis problems that actively deceive an adversary about a system's underlying rewards. The first form of deception is ``diversionary'', and it leads an adversary to draw any false conclusion about what the system's reward function is. The second form of deception is ``targeted'', and it leads an adversary to draw a specific false conclusion about what the system's reward function is. We then show how each form of deception can be implemented in policy optimization problems, and we analytically bound the loss in total accumulated reward that is induced by deception. Next, we evaluate these developments in a multi-agent sequential decision-making problem with one real agent and multiple decoys. We show that diversionary deception can cause the adversary to believe that the most important agent is the least important, while attaining a total accumulated reward that is $98.83\%$ of its optimal, non-deceptive value. Similarly, we show that targeted deception can make any decoy appear to be the most important agent, while still attaining a total accumulated reward that is $99.25\%$ of its optimal, non-deceptive value.
The combinatorial multi-armed bandit (CMAB) is a fundamental sequential decision-making framework, extensively studied over the past decade. However, existing work primarily focuses on the online setting, overlooking the substantial costs of online interactions and the readily available offline datasets. To overcome these limitations, we introduce Off-CMAB, the first offline learning framework for CMAB. Central to our framework is the combinatorial lower confidence bound (CLCB) algorithm, which combines pessimistic reward estimations with combinatorial solvers. To characterize the quality of offline datasets, we propose two novel data coverage conditions and prove that, under these conditions, CLCB achieves a near-optimal suboptimality gap, matching the theoretical lower bound up to a logarithmic factor. We validate Off-CMAB through practical applications, including learning to rank, large language model (LLM) caching, and social influence maximization, showing its ability to handle nonlinear reward functions, general feedback models, and out-of-distribution action samples that excludes optimal or even feasible actions. Extensive experiments on synthetic and real-world datasets further highlight the superior performance of CLCB.
Contextual sequential decision-making problems play a crucial role in machine learning, encompassing a wide range of downstream applications such as bandits, sequential hypothesis testing and online risk control. These applications often require different statistical measures, including expectation, variance and quantiles. In this paper, we provide a universal admissible algorithm framework for dealing with all kinds of contextual online decision-making problems that directly learns the whole underlying unknown distribution instead of focusing on individual statistics. This is much more difficult because the dimension of the regression is uncountably infinite, and any existing linear contextual bandits algorithm will result in infinite regret. To overcome this issue, we propose an efficient infinite-dimensional functional regression oracle for contextual cumulative distribution functions (CDFs), where each data point is modeled as a combination of context-dependent CDF basis functions. Our analysis reveals that the decay rate of the eigenvalue sequence of the design integral operator governs the regression error rate and, consequently, the utility regret rate. Specifically, when the eigenvalue sequence exhibits a polynomial decay of order $\frac{1}{\gamma}\ge 1$, the utility regret is bounded by $\tilde{\mathcal{O}}\Big(T^{\frac{3\gamma+2}{2(\gamma+2)}}\Big)$. By setting $\gamma=0$, this recovers the existing optimal regret rate for contextual bandits with finite-dimensional regression and is optimal under a stronger exponential decay assumption. Additionally, we provide a numerical method to compute the eigenvalue sequence of the integral operator, enabling the practical implementation of our framework.
Hyperdimensional Computing (HDC), also known as Vector Symbolic Architectures, is a computing paradigm that combines the strengths of symbolic reasoning with the efficiency and scalability of distributed connectionist models in artificial intelligence. HDC has recently emerged as a promising alternative for performing learning tasks in resource-constrained environments thanks to its energy and computational efficiency, inherent parallelism, and resilience to noise and hardware faults. This work introduces the Hyperdimensional Contextual Bandits (HD-CB): the first exploration of HDC to model and automate sequential decision-making Contextual Bandits (CB) problems. The proposed approach maps environmental states in a high-dimensional space and represents each action with dedicated hypervectors (HVs). At each iteration, these HVs are used to select the optimal action for the given context and are updated based on the received reward, replacing computationally expensive ridge regression procedures required by traditional linear CB algorithms with simple, highly parallel vector operations. We propose four HD-CB variants, demonstrating their flexibility in implementing different exploration strategies, as well as techniques to reduce memory overhead and the number of hyperparameters. Extensive simulations on synthetic datasets and a real-world benchmark reveal that HD-CB consistently achieves competitive or superior performance compared to traditional linear CB algorithms, while offering faster convergence time, lower computational complexity, improved scalability, and high parallelism.
Behavior cloning has shown success in many sequential decision-making tasks by learning from expert demonstrations, yet they can be very sample inefficient and fail to generalize to unseen scenarios. One approach to these problems is to introduce general domain knowledge, such that the policy can focus on the essential features and may generalize to unseen states by applying that knowledge. Although this knowledge is easy to acquire from the experts, it is hard to be combined with learning from individual examples due to the lack of semantic structure in neural networks and the time-consuming nature of feature engineering. To enable learning from both general knowledge and specific demonstration trajectories, we use a large language model's coding capability to instantiate a policy structure based on expert domain knowledge expressed in natural language and tune the parameters in the policy with demonstrations. We name this approach the Knowledge Informed Model (KIM) as the structure reflects the semantics of expert knowledge. In our experiments with lunar lander and car racing tasks, our approach learns to solve the tasks with as few as 5 demonstrations and is robust to action noise, outperforming the baseline model without domain knowledge. This indicates that with the help of large language models, we can incorporate domain knowledge into the structure of the policy, increasing sample efficiency for behavior cloning.