Many practical reinforcement learning environments have a discrete factored action space that induces a large combinatorial set of actions, thereby posing significant challenges. Existing approaches leverage the regular structure of the action space and resort to a linear decomposition of Q-functions, which avoids enumerating all combinations of factored actions. In this paper, we consider Q-functions defined over a lower dimensional projected subspace of the original action space, and study the condition for the unbiasedness of decomposed Q-functions using causal effect estimation from the no unobserved confounder setting in causal statistics. This leads to a general scheme which we call action decomposed reinforcement learning that uses the projected Q-functions to approximate the Q-function in standard model-free reinforcement learning algorithms. The proposed approach is shown to improve sample complexity in a model-based reinforcement learning setting. We demonstrate improvements in sample efficiency compared to state-of-the-art baselines in online continuous control environments and a real-world offline sepsis treatment environment.

Developing value-aligned agents is a complex undertaking and an ongoing challenge in the field of AI. Indeed, designing Large Language Models (LLMs) that can balance multiple possibly conflicting moral values based on the context is a problem of paramount importance. In this paper, we propose a system that performs contextual value alignment based on contextual aggregation of possible responses. This aggregation is achieved by integrating a subset of possible LLM responses that are best suited to a user's input while taking into account features extracted about the user's moral preferences. The proposed system trained using the Moral Integrity Corpus displays better alignment to human values than state-of-the-art baselines.

Erik Miehling, Michael Desmond, Karthikeyan Natesan Ramamurthy, Elizabeth M. Daly, Kush R. Varshney, Eitan Farchi, Pierre Dognin, Jesus Rios, Djallel Bouneffouf, Miao Liu, Prasanna Sattigeri. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

Causal discovery in the form of a directed acyclic graph (DAG) for dynamic time series data has been widely studied in various applications. Much of the existing work has focused on observational, offline, and/or stationary settings. In this work, we propose a dynamic DAG discovery algorithm, Meta-D$^2$AG, based on online meta-learning. Meta-D$^2$AG is designed to learn dynamic DAG structures from potentially nonlinear and non-stationary times series datasets, accounting for changes in both parameters and graph structures. Notably, Meta-D$^2$AG explicitly treats data collected at different time points with distribution shifts as distinct domains, which is assumed to occur as a result of external interventions. Moreover, Meta-D$^2$AG contains a new online meta-learning framework to take advantage of the temporal transition among existing domains such that it can quickly adapt to new domains with few measurements. A first-order optimization approach is utilized to efficiently solve the meta-learning framework, and theoretical analysis establishes the identifiability conditions and the convergence of the learning process. We demonstrate the promising performance of our method through better accuracy and sample efficiency on benchmark datasets against state-of-the-art baselines.

Many machine learning problems today have multiple objective functions, which are often tackled by the multi-objective learning (MOL) framework. Albeit many encouraging results are obtained by MOL algorithms, a recent theoretical study [1] revealed that these gradient-based MOL methods (e.g., MGDA, CAGrad) all reflect an inherent trade-off between optimization convergence speeds and conflict-avoidance abilities. To this end, we develop an improved stochastic variance-reduced multi-objective gradient correction method for MOL, achieving the ${\mathcal{O}}\left({{\varepsilon ^{ - 1.5}}}\right)$ sample complexity. In addition, our proposed method simultaneously improves the theoretical guarantees for conflict avoidance and convergence rate compared to prior stochastic gradient-based MOL methods in the non-convex setting. We further validate the effectiveness of the proposed method empirically using popular multi-task learning (MTL) benchmarks.

This paper studies the transfer reinforcement learning (RL) problem where multiple RL problems have different reward functions but share the same underlying transition dynamics. In this setting, the Q-function of each RL problem (task) can be decomposed into a successor feature (SF) and a reward mapping: the former characterizes the transition dynamics, and the latter characterizes the task-specific reward function. This Q-function decomposition, coupled with a policy improvement operator known as generalized policy improvement (GPI), reduces the sample complexity of finding the optimal Q-function, and thus the SF & GPI framework exhibits promising empirical performance compared to traditional RL methods like Q-learning. However, its theoretical foundations remain largely unestablished, especially when learning the successor features using deep neural networks (SF-DQN). This paper studies the provable knowledge transfer using SFs-DQN in transfer RL problems. We establish the first convergence analysis with provable generalization guarantees for SF-DQN with GPI. The theory reveals that SF-DQN with GPI outperforms conventional RL approaches, such as deep Q-network, in terms of both faster convergence rate and better generalization. Numerical experiments on real and synthetic RL tasks support the superior performance of SF-DQN & GPI, aligning with our theoretical findings.

Hierarchical reinforcement learning has focused on discovering temporally extended actions, such as options, that can provide benefits in problems requiring extensive exploration. One promising approach that learns these options end-to-end is the option-critic (OC) framework. We examine and show in this paper that OC does not decompose a problem into simpler sub-problems, but instead increases the size of the search over policy space with each option considering the entire state space during learning. This issue can result in practical limitations of this method, including sample inefficient learning. To address this problem, we introduce Context-Specific Representation Abstraction for Deep Option Learning (CRADOL), a new framework that considers both temporal abstraction and context-specific representation abstraction to effectively reduce the size of the search over policy space. Specifically, our method learns a factored belief state representation that enables each option to learn a policy over only a subsection of the state space. We test our method against hierarchical, non-hierarchical, and modular recurrent neural network baselines, demonstrating significant sample efficiency improvements in challenging partially observable environments.

Causal discovery in the form of a directed acyclic graph (DAG) for time series data has been widely studied in various domains. The resulting DAG typically represents a dynamic Bayesian network (DBN), capturing both the instantaneous and time-delayed relationships among variables of interest. We propose a new algorithm, IDYNO, to learn the DAG structure from potentially nonlinear times series data by using a continuous optimization framework that includes a recent formulation for continuous acyclicity constraint. The proposed algorithm is designed to handle both observational and interventional time series data. We demonstrate the promising performance of our method on synthetic benchmark datasets against state-of-the-art baselines. In addition, we show that the proposed method can more accurately learn the underlying structure of a sequential decision model, such as a Markov decision process, with a fixed policy in typical continuous control tasks.

The main challenge of multiagent reinforcement learning is the difficulty of learning useful policies in the presence of other simultaneously learning agents whose changing behaviors jointly affect the environment's transition and reward dynamics. An effective approach that has recently emerged for addressing this non-stationarity is for each agent to anticipate the learning of other agents and influence the evolution of future policies towards desirable behavior for its own benefit. Unfortunately, previous approaches for achieving this suffer from myopic evaluation, considering only a finite number of policy updates. As such, these methods can only influence transient future policies rather than achieving the promise of scalable equilibrium selection approaches that influence the behavior at convergence. In this paper, we propose a principled framework for considering the limiting policies of other agents as time approaches infinity. Specifically, we develop a new optimization objective that maximizes each agent's average reward by directly accounting for the impact of its behavior on the limiting set of policies that other agents will converge to. Our paper characterizes desirable solution concepts within this problem setting and provides practical approaches for optimizing over possible outcomes. As a result of our farsighted objective, we demonstrate better long-term performance than state-of-the-art baselines across a suite of diverse multiagent benchmark domains.

In many real-world applications of multi-armed bandit problems, both rewards and contexts are often influenced by confounding latent variables which evolve stochastically over time. While the observed contexts and rewards are nonlinearly related, we show that prior knowledge of latent causal structure can be used to reduce the problem to the linear bandit setting. We develop two algorithms, Latent Linear Thompson Sampling (L2TS) and Latent Linear UCB (L2UCB), which use online EM algorithms for hidden Markov models to learn the latent transition model and maintain a posterior belief over the latent state, and then use the resulting posteriors as context features in a linear bandit problem. We upper bound the error in reward estimation in the presence of a dynamical latent state, and derive a novel problem-dependent regret bound for linear Thompson sampling with non-stationarity and unconstrained reward distributions, which we apply to L2TS under certain conditions. Finally, we demonstrate the superiority of our algorithms over related bandit algorithms through experiments.

Training agents to learn efficiently in multi-agent environments can benefit from the explicit modelling of other agent's beliefs, especially in complex limited-information games such as the Hanabi card game. However, generalization is also highly relevant to performance in these games, though model comparisons at large training timescales can be difficult. In this work, we address this by introducing a novel model trained using a sleep metaphor on a reduced complexity version of the Hanabi game. This sleep metaphor consists an altered training regiment, as well as an information-theoretic constraint on the agent's policy. Results from experimentation demonstrate improved performance through this sleep-metaphor method, and provide a promising motivation for using similar techniques in more complex methods that incorporate explicit models of other agent's beliefs.

This paper explores information-theoretic constraints on methods for multi-agent reinforcement learning (MARL) in mixed cooperative and competitive games. Within this domain, decentralized training has been employed to increase learning sample efficiency. However, these approaches do not explicitly discourage complex policies, which can lead to overfitting. To address this, we apply an information theoretic constraint onto agents' policies that discourages overly complex behaviour when it is not associated with a significant increase in reward. A second challenge in MARL is the non-stationarity of the environment introduced by other agents' changing policies. Previous methods in MARL have sought to reduce the impact of non-stationarity by inferring other agents' policies, but this can lead to over-fitting to previously observed behaviour. To avoid this, a similar information-theoretic constraint is applied onto the inference of other agents' policies, resulting in a more robust estimate. We evaluate the effects of these information-theoretic constraints on a test suite of multi-agent games, and report an overall improvement in performance, with greater improvements found in competitive domains compared to cooperative games.

A fundamental challenge in multiagent reinforcement learning is to learn beneficial behaviors in a shared environment with other simultaneously learning agents. In particular, each agent perceives the environment as effectively non-stationary due to the changing policies of other agents. Moreover, each agent is itself constantly learning, leading to natural non-stationarity in the distribution of experiences encountered. In this paper, we propose a novel meta-multiagent policy gradient theorem that directly accounts for the non-stationary policy dynamics inherent to multiagent learning settings. This is achieved by modeling our gradient updates to consider both an agent’s own non-stationary policy dynamics and the non-stationary policy dynamics of other agents in the environment. We show that our theoretically grounded approach provides a general solution to the multiagent learning problem, which inherently comprises all key aspects of previous state of the art approaches on this topic. We test our method on a diverse suite of multiagent benchmarks and demonstrate a more efficient ability to adapt to new agents as they learn than baseline methods across the full spectrum of mixed incentive, competitive, and cooperative domains.

The options framework is a popular approach for building temporally extended actions in reinforcement learning. In particular, the option-critic architecture provides general purpose policy gradient theorems for learning actions from scratch that are extended in time. However, past work makes the key assumption that each of the components of option-critic has independent parameters. In this work we note that while this key assumption of the policy gradient theorems of option-critic holds in the tabular case, it is always violated in practice for the deep function approximation setting. We thus reconsider this assumption and consider more general extensions of option-critic and hierarchical option-critic training that optimize for the full architecture with each update. It turns out that not assuming parameter independence challenges a belief in prior work that training the policy over options can be disentangled from the dynamics of the underlying options. In fact, learning can be sped up by focusing the policy over options on states where options are actually likely to terminate. We put our new algorithms to the test in application to sample efficient learning of Atari games, and demonstrate significantly improved stability and faster convergence when learning long options. 1

Collective human knowledge has clearly benefited from the fact that innovations by individuals are taught to others through communication. Similar to human social groups, agents in distributed learning systems would likely benefit from communication to share knowledge and teach skills. The problem of teaching to improve agent learning has been investigated by prior works, but these approaches make assumptions that prevent application of teaching to general multiagent problems, or require domain expertise for problems they can apply to. This learning to teach problem has inherent complexities related to measuring long-term impacts of teaching that compound the standard multiagent coordination challenges. In contrast to existing works, this paper presents the first general framework and algorithm for intelligent agents to learn to teach in a multiagent environment. Our algorithm, Learning to Coordinate and Teach Reinforcement (LeCTR), addresses peer-to-peer teaching in cooperative multiagent reinforcement learning. Each agent in our approach learns both when and what to advise, then uses the received advice to improve local learning. Importantly, these roles are not fixed; these agents learn to assume the role of student and/or teacher at the appropriate moments, requesting and providing advice in order to improve teamwide performance and learning. Empirical comparisons against state-of-the-art teaching methods show that our teaching agents not only learn significantly faster, but also learn to coordinate in tasks where existing methods fail.

Building systems that autonomously create temporal abstractions from data is a key challenge in scaling learning and planning in reinforcement learning. One popular approach for addressing this challenge is the options framework (Sutton et al., 1999). However, only recently in (Bacon et al., 2017) was a policy gradient theorem derived for online learning of general purpose options in an end to end fashion. In this work, we extend previous work on this topic that only focuses on learning a two-level hierarchy including options and primitive actions to enable learning simultaneously at multiple resolutions in time. We achieve this by considering an arbitrarily deep hierarchy of options where high level temporally extended options are composed of lower level options with finer resolutions in time. We extend results from (Bacon et al., 2017) and derive policy gradient theorems for a deep hierarchy of options. Our proposed hierarchical option-critic architecture is capable of learning internal policies, termination conditions, and hierarchical compositions over options without the need for any intrinsic rewards or subgoals. Our empirical results in both discrete and continuous environments demonstrate the efficiency of our framework.

Finding feasible, collision-free paths for multiagent systems can be challenging, particularly in non-communicating scenarios where each agent's intent (e.g. goal) is unobservable to the others. In particular, finding time efficient paths often requires anticipating interaction with neighboring agents, the process of which can be computationally prohibitive. This work presents a decentralized multiagent collision avoidance algorithm based on a novel application of deep reinforcement learning, which effectively offloads the online computation (for predicting interaction patterns) to an offline learning procedure. Specifically, the proposed approach develops a value network that encodes the estimated time to the goal given an agent's joint configuration (positions and velocities) with its neighbors. Use of the value network not only admits efficient (i.e., real-time implementable) queries for finding a collision-free velocity vector, but also considers the uncertainty in the other agents' motion. Simulation results show more than 26% improvement in paths quality (i.e., time to reach the goal) when compared with optimal reciprocal collision avoidance (ORCA), a state-of-the-art collision avoidance strategy.

This paper presents the first ever approach for solving continuous-observation Decentralized Partially Observable Markov Decision Processes (Dec-POMDPs) and their semi-Markovian counterparts, Dec-POSMDPs. This contribution is especially important in robotics, where a vast number of sensors provide continuous observation data. A continuous-observation policy representation is introduced using Stochastic Kernel-based Finite State Automata (SK-FSAs). An SK-FSA search algorithm titled Entropy-based Policy Search using Continuous Kernel Observations (EPSCKO) is introduced and applied to the first ever continuous-observation Dec-POMDP/Dec-POSMDP domain, where it significantly outperforms state-of-the-art discrete approaches. This methodology is equally applicable to Dec-POMDPs and Dec-POSMDPs, though the empirical analysis presented focuses on Dec-POSMDPs due to their higher scalability. To improve convergence, an entropy injection policy search acceleration approach for both continuous and discrete observation cases is also developed and shown to improve convergence rates without degrading policy quality.

Robust environment perception is essential for decision-making on robots operating in complex domains. Intelligent task execution requires principled treatment of uncertainty sources in a robot's observation model. This is important not only for low-level observations (e.g., accelerom-eter data), but also for high-level observations such as semantic object labels. This paper formalizes the concept of macro-observations in Decentralized Partially Observable Semi-Markov Decision Processes (Dec-POSMDPs), allowing scalable semantic-level multi-robot decision making. A hierarchical Bayesian approach is used to model noise statistics of low-level classifier outputs, while simultaneously allowing sharing of domain noise characteristics between classes. Classification accuracy of the proposed macro-observation scheme, called Hierarchical Bayesian Noise Inference (HBNI), is shown to exceed existing methods. The macro-observation scheme is then integrated into a Dec-POSMDP planner, with hardware experiments running onboard a team of dynamic quadrotors in a challenging domain where noise-agnostic filtering fails. To the best of our knowledge, this is the first demonstration of a real-time, convolutional neural net-based classification framework running fully onboard a team of quadrotors in a multi-robot decision-making domain.

For robotic vehicles to navigate safely and efficiently in pedestrian-rich environments, it is important to model subtle human behaviors and navigation rules (e.g., passing on the right). However, while instinctive to humans, socially compliant navigation is still difficult to quantify due to the stochasticity in people's behaviors. Existing works are mostly focused on using feature-matching techniques to describe and imitate human paths, but often do not generalize well since the feature values can vary from person to person, and even run to run. This work notes that while it is challenging to directly specify the details of what to do (precise mechanisms of human navigation), it is straightforward to specify what not to do (violations of social norms). Specifically, using deep reinforcement learning, this work develops a time-efficient navigation policy that respects common social norms. The proposed method is shown to enable fully autonomous navigation of a robotic vehicle moving at human walking speed in an environment with many pedestrians.

This paper presents a data-driven approach for multi-robot coordination in partially-observable domains based on Decentralized Partially Observable Markov Decision Processes (Dec-POMDPs) and macro-actions (MAs). Dec-POMDPs provide a general framework for cooperative sequential decision making under uncertainty and MAs allow temporally extended and asynchronous action execution. To date, most methods assume the underlying Dec-POMDP model is known a priori or a full simulator is available during planning time. Previous methods which aim to address these issues suffer from local optimality and sensitivity to initial conditions. Additionally, few hardware demonstrations involving a large team of heterogeneous robots and with long planning horizons exist. This work addresses these gaps by proposing an iterative sampling based Expectation-Maximization algorithm (iSEM) to learn polices using only trajectory data containing observations, MAs, and rewards. Our experiments show the algorithm is able to achieve better solution quality than the state-of-the-art learning-based methods. We implement two variants of multi-robot Search and Rescue (SAR) domains (with and without obstacles) on hardware to demonstrate the learned policies can effectively control a team of distributed robots to cooperate in a partially observable stochastic environment.

Decentralized partially observable Markov decision processes (Dec-POMDPs) provide a general framework for multiagent sequential decision-making under uncertainty. Although Dec-POMDPs are typically intractable to solve for real-world problems, recent research on macro-actions (i.e., temporally-extended actions) has significantly increased the size of problems that can be solved. However, current methods assume the underlying Dec-POMDP model is known a priori or a full simulator is available during planning time. To accommodate more realistic scenarios, when such information is not available, this paper presents a policy-based reinforcement learning approach, which learns the agent policies based solely on trajectories generated by previous interaction with the environment (e.g., demonstrations). We show that our approach is able to generate valid macro-action controllers and develop an expectationmaximization (EM) algorithm (called Policy-based EM or PoEM), which has convergence guarantees for batch learning. Our experiments show PoEM is a scalable learning method that can learn optimal policies and improve upon hand-coded “expert” solutions.

Developing accurate models and efficient representations of multivariate trajectories is important for understanding the behavior patterns of mobile agents. This work presents a dictionary learning algorithm for developing a part-based trajectory representation, which combines merits of the existing Markovian-based and clustering-based approaches. In particular, this work presents the augmented semi-nonnegative sparse coding (ASNSC) algorithm for solving a constrained dictionary learning problem, and shows that the proposed method would converge to a local optimum given a convexity condition. We consider a trajectory modeling application, in which the learned dictionary atoms correspond to local motion patterns. Classical semi-nonnegative sparse coding approaches would add dictionary atoms with opposite signs to reduce the representational error, which can lead to learning noisy dictionary atoms that correspond poorly to local motion patterns. ASNSC addresses this problem and learns a concise set of intuitive motion patterns. ASNSC shows significant improvement over existing trajectory modeling methods in both prediction accuracy and computational time, as revealed by extensive numerical analysis on real datasets.

Many robotic applications require repeated, on-demand motion planning in mapped environments. In addition, the presence of other dynamic agents, such as people, often induces frequent, dynamic changes in the environment. Having a potential function that encodes pairwise cost-to-go can be useful for improving the computational speed of finding feasible paths, and for guiding local searches around dynamic obstacles. However, since storing pairwise potential can be impractical given the O(|V|2) memory requirement, existing work often needs to compute a potential function for each query to a new goal, which would require a substantial online computation. This work addresses the problem by using diffusion maps, a machine learning algorithm, to learn the map's geometry and develop a memory-efficient parametrization (O(|V|)) of pairwise potentials. Specially, each state in the map is transformed to a diffusion coordinate, in which pairwise Euclidean distance is shown to be a meaningful similarity metric. We develop diffusion-based motion planning algorithms and, through extensive numerical evaluation, show that the proposed algorithms find feasible paths of similar quality with orders of magnitude improvement in computational speed compared with single-query methods. The proposed algorithms are implemented on hardware to enable real-time autonomous navigation in an indoor environment with frequent interactions with pedestrians.

This paper presents a novel algorithm, based upon the dependent Dirichlet process mixture model (DDPMM), for clustering batch-sequential data containing an unknown number of evolving clusters. The algorithm is derived via a low-variance asymptotic analysis of the Gibbs sampling algorithm for the DDPMM, and provides a hard clustering with convergence guarantees similar to those of the k-means algorithm. Empirical results from a synthetic test with moving Gaussian clusters and a test with real ADS-B aircraft trajectory data demonstrate that the algorithm requires orders of magnitude less computational time than contemporary probabilistic and hard clustering algorithms, while providing higher accuracy on the examined datasets.