The improvement of economic policymaking presents an opportunity for broad societal benefit, a notion that has inspired research towards AI-driven policymaking tools. AI policymaking holds the potential to surpass human performance through the ability to process data quickly at scale. However, existing RL-based methods exhibit sample inefficiency, and are further limited by an inability to flexibly incorporate nuanced information into their decision-making processes. Thus, we propose a novel method in which we instead utilize pre-trained Large Language Models (LLMs), as sample-efficient policymakers in socially complex multi-agent reinforcement learning (MARL) scenarios. We demonstrate significant efficiency gains, outperforming existing methods across three environments. Our code is available at https://github.com/hegasz/large-legislative-models.

Recent advancements in test time compute, particularly through the use of verifier models, have significantly enhanced the reasoning capabilities of Large Language Models (LLMs). This generator-verifier approach closely resembles the actor-critic framework in reinforcement learning (RL). However, current verifier models in LLMs often rely on supervised fine-tuning without temporal difference learning such as Q-learning. This paper introduces VerifierQ, a novel approach that integrates Offline Q-learning into LLM verifier models. We address three key challenges in applying Q-learning to LLMs: (1) handling utterance-level Markov Decision Processes (MDPs), (2) managing large action spaces, and (3) mitigating overestimation bias. VerifierQ introduces a modified Bellman update for bounded Q-values, incorporates Implicit Q-learning (IQL) for efficient action space management, and integrates a novel Conservative Q-learning (CQL) formulation for balanced Q-value estimation. Our method enables parallel Q-value computation and improving training efficiency. While recent work has explored RL techniques like MCTS for generators, VerifierQ is among the first to investigate the verifier (critic) aspect in LLMs through Q-learning. This integration of RL principles into verifier models complements existing advancements in generator techniques, potentially enabling more robust and adaptive reasoning in LLMs. Experimental results on mathematical reasoning tasks demonstrate VerifierQ's superior performance compared to traditional supervised fine-tuning approaches, with improvements in efficiency, accuracy and robustness. By enhancing the synergy between generation and evaluation capabilities, VerifierQ contributes to the ongoing evolution of AI systems in addressing complex cognitive tasks across various domains.

In the realm of neuroscience, it has been observed that biological agents often experience a diminishing ability to adapt over time, analogous to the gradual solidification of neural pathways in the brain (Livingston, 1966). This phenomenon, typically known as the loss of plasticity (Mateos-Aparicio and Rodríguez-Moreno, 2019), significantly affects an agent's capacity to learn continually, especially when agents learn by trial and error in deep reinforcement learning (deep RL) due to the nonstationarity nature. The declining adaptability throughout the learning process can severely hinder the agent's ability to effectively learn and respond to complex or non-stationary scenarios (Abbas et al., 2023). This limitation presents a fundamental obstacle to achieving sustained learning and adaptability in artificial agents, which echoes the plasticity-stability dilemma (Abraham and Robins, 2005) observed in biological neural networks. There have been several recent studies highlighting a significant loss of plasticity in deep RL (Kumar et al., 2020, Lyle et al., 2022), which substantially restricts the agent's ability to learn from subsequent experiences (Lyle et al., 2023, Ma et al., 2023). The identification of primacy bias (Nikishin et al., 2022) further illustrates how agents may become overfitted to early experiences, which inhibits learning from subsequent new data. The consequences of plasticity loss further impede deep RL in continual learning scenarios, where the agent struggles to sequentially learn across a series of different tasks (Dohare et al., 2024). 1

Hierarchical policies enable strong performance in many sequential decision-making problems, such as those with high-dimensional action spaces, those requiring long-horizon planning, and settings with sparse rewards. However, learning hierarchical policies from static offline datasets presents a significant challenge. Crucially, actions taken by higher-level policies may not be directly observable within hierarchical controllers, and the offline dataset might have been generated using a different policy structure, hindering the use of standard offline learning algorithms. In this work, we propose OHIO: a framework for offline reinforcement learning (RL) of hierarchical policies. Our framework leverages knowledge of the policy structure to solve the inverse problem, recovering the unobservable high-level actions that likely generated the observed data under our hierarchical policy. This approach constructs a dataset suitable for off-the-shelf offline training. We demonstrate our framework on robotic and network optimization problems and show that it substantially outperforms end-to-end RL methods and improves robustness. We investigate a variety of instantiations of our framework, both in direct deployment of policies trained offline and when online fine-tuning is performed.

In sequential decision-making (SDM) tasks, methods like reinforcement learning (RL) and heuristic search have made notable advances in specific cases. However, they often require extensive exploration and face challenges in generalizing across diverse environments due to their limited grasp of the underlying decision dynamics. In contrast, large language models (LLMs) have recently emerged as powerful general-purpose tools, due to their capacity to maintain vast amounts of domainspecific knowledge. To harness this rich prior knowledge for efficiently solving complex SDM tasks, we propose treating LLMs as prior action distributions and integrating them into RL frameworks through Bayesian inference methods, making use of variational inference and direct posterior sampling. The proposed approaches facilitate the seamless incorporation of fixed LLM priors into both policy-based and value-based RL frameworks. Our experiments show that incorporating LLMbased action priors significantly reduces exploration and optimization complexity, substantially improving sample efficiency compared to traditional RL techniques, e.g., using LLM priors decreases the number of required samples by over 90% in offline learning scenarios. Traditional approaches to SDM, such as optimal control (Garcia et al., 1989), heuristic search (Świechowski et al., 2023) and reinforcement learning (RL) (Mnih, 2013), have seen substantial success. Notably, AlphaGo (Silver et al., 2016) and AlphaStar (Vinyals et al., 2019), both based on deep reinforcement learning (DRL), have achieved human-level proficiency in the games of Go and StarCraft II, respectively. However, these methods still suffer from high computational complexity, along with poor generalizability and limited applicability across diverse domains (Dulac-Arnold et al., 2015; Cobbe et al., 2019). Recently, Large Language Models (LLMs) have emerged as effective tools for tackling diverse general-purpose tasks, such as in dialogue systems (Brooks et al., 2023), decision-making (Zhao et al., 2024a), and mathematical reasoning (Imani et al., 2023).

Hierarchical Reinforcement Learning (HRL) effectively tackles complex tasks by decomposing them into structured policies. However, HRL agents often face challenges with efficient exploration and rapid adaptation. To address this, we integrate meta-learning into HRL to enhance the agent's ability to learn and adapt hierarchical policies swiftly. Our approach employs meta-learning for rapid task adaptation based on prior experience, while intrinsic motivation mechanisms encourage efficient exploration by rewarding novel state visits. Specifically, our agent uses a high-level policy to select among multiple low-level policies operating within custom grid environments. We utilize gradient-based meta-learning with differentiable inner-loop updates, enabling optimization across a curriculum of increasingly difficult tasks. Experimental results demonstrate that our meta-learned hierarchical agent significantly outperforms traditional HRL agents without meta-learning and intrinsic motivation. The agent exhibits accelerated learning, higher cumulative rewards, and improved success rates in complex grid environments. These findings suggest that integrating meta-learning with HRL, alongside curriculum learning and intrinsic motivation, substantially enhances the agent's capability to handle complex tasks.

Representation learning and unsupervised skill discovery can allow robots to acquire diverse and reusable behaviors without the need for task-specific rewards. In this work, we use unsupervised reinforcement learning to learn a latent representation by maximizing the mutual information between skills and states subject to a distance constraint. Our method improves upon prior constrained skill discovery methods by replacing the latent transition maximization with a norm-matching objective. This not only results in a much a richer state space coverage compared to baseline methods, but allows the robot to learn more stable and easily controllable locomotive behaviors. We successfully deploy the learned policy on a real ANYmal quadruped robot and demonstrate that the robot can accurately reach arbitrary points of the Cartesian state space in a zero-shot manner, using only an intrinsic skill discovery and standard regularization rewards.

A crucial capability of Machine Learning models in real-world applications is the ability to continuously learn new tasks. This adaptability allows them to respond to potentially inevitable shifts in the data-generating distribution over time. However, in Continual Learning (CL) settings, models often struggle to balance learning new tasks (plasticity) with retaining previous knowledge (memory stability). Consequently, they are susceptible to Catastrophic Forgetting, which degrades performance and undermines the reliability of deployed systems. Variational Continual Learning methods tackle this challenge by employing a learning objective that recursively updates the posterior distribution and enforces it to stay close to the latest posterior estimate. Nonetheless, we argue that these methods may be ineffective due to compounding approximation errors over successive recursions. To mitigate this, we propose new learning objectives that integrate the regularization effects of multiple previous posterior estimations, preventing individual errors from dominating future posterior updates and compounding over time. We reveal insightful connections between these objectives and Temporal-Difference methods, a popular learning mechanism in Reinforcement Learning and Neuroscience. We evaluate the proposed objectives on challenging versions of popular CL benchmarks, demonstrating that they outperform standard Variational CL methods and non-variational baselines, effectively alleviating Catastrophic Forgetting.

Optimization of user association in a densely deployed heterogeneous cellular network is usually challenging and even more complicated due to the dynamic nature of user mobility and fluctuation in user counts. While deep reinforcement learning (DRL) emerges as a promising solution, its application in practice is hindered by high trial-and-error costs in real world and unsatisfactory physical network performance during training. In addition, existing DRL-based user association methods are usually only applicable to scenarios with a fixed number of users due to convergence and compatibility challenges. In this paper, we propose a parallel digital twin (DT)-driven DRL method for user association and load balancing in networks with both dynamic user counts, distribution, and mobility patterns. Our method employs a distributed DRL strategy to handle varying user numbers and exploits a refined neural network structure for faster convergence. To address these DRL training-related challenges, we devise a high-fidelity DT construction technique, featuring a zero-shot generative user mobility model, named Map2Traj, based on a diffusion model. Map2Traj estimates user trajectory patterns and spatial distributions solely from street maps. Armed with this DT environment, DRL agents are enabled to be trained without the need for interactions with the physical network. To enhance the generalization ability of DRL models for dynamic scenarios, a parallel DT framework is further established to alleviate strong correlation and non-stationarity in single-environment training and improve the training efficiency. Numerical results show that the proposed parallel DT-driven DRL method achieves closely comparable performance to real environment training, and even outperforms those trained in a single real-world environment with nearly 20% gain in terms of cell-edge user performance.

Multi-agent reinforcement learning (MARL) presents unique challenges as agents learn strategies through experiences. Gradient-based methods are often sensitive to hyperparameter selection and initial random seed variations. Concurrently, significant advances have been made in solving Variational Inequalities (VIs) which include equilibrium-finding problems particularly in addressing the non-converging rotational dynamics that impede convergence of traditional gradient based optimization methods. This paper explores the potential of leveraging VI-based techniques to improve MARL training. Specifically, we study the performance of VI method namely, Nested-Lookahead VI (nLA-VI) and Extragradient (EG) in enhancing the multi-agent deep deterministic policy gradient (MADDPG) algorithm. We present a VI reformulation of the actor-critic algorithm for both single- and multi-agent settings. We introduce three algorithms that use nLA-VI, EG, and a combination of both, named LA-MADDPG, EG-MADDPG, and LA-EG-MADDPG, respectively. Our empirical results demonstrate that these VI-based approaches yield significant performance improvements in benchmark environments, such as the zero-sum games: rock-paper-scissors and matching pennies, where equilibrium strategies can be quantitatively assessed, and the Multi-Agent Particle Environment: Predator prey benchmark, where VI-based methods also yield balanced participation of agents from the same team.