Large Language Models (LLMs) still struggle with complex reasoning tasks. Motivated by the society of minds (Minsky, 1988), we propose ReConcile, a multi-model multi-agent framework designed as a round table conference among diverse LLM agents to foster diverse thoughts and discussion for improved consensus. ReConcile enhances the reasoning capabilities of LLMs by holding multiple rounds of discussion, learning to convince other agents to improve their answers, and employing a confidence-weighted voting mechanism. In each round, ReConcile initiates discussion between agents via a 'discussion prompt' that consists of (a) grouped answers and explanations generated by each agent in the previous round, (b) their uncertainties, and (c) demonstrations of answer-rectifying human explanations, used for convincing other agents. This discussion prompt enables each agent to revise their responses in light of insights from other agents. Once a consensus is reached and the discussion ends, ReConcile determines the final answer by leveraging the confidence of each agent in a weighted voting scheme. We implement ReConcile with ChatGPT, Bard, and Claude2 as the three agents. Our experimental results on various benchmarks demonstrate that ReConcile significantly enhances the reasoning performance of the agents (both individually and as a team), surpassing prior single-agent and multi-agent baselines by 7.7% and also outperforming GPT-4 on some of these datasets. We also experiment with GPT-4 itself as one of the agents in ReConcile and demonstrate that its initial performance also improves by absolute 10.0% through discussion and feedback from other agents. Finally, we also analyze the accuracy after every round and observe that ReConcile achieves better and faster consensus between agents, compared to a multi-agent debate baseline. Our code is available at: https://github.com/dinobby/ReConcile

Even though Google Research Football (GRF) was initially benchmarked and studied as a single-agent environment in its original paper, recent years have witnessed an increasing focus on its multi-agent nature by researchers utilizing it as a testbed for Multi-Agent Reinforcement Learning (MARL). However, the absence of standardized environment settings and unified evaluation metrics for multi-agent scenarios hampers the consistent understanding of various studies. Furthermore, the challenging 5-vs-5 and 11-vs-11 full-game scenarios have received limited thorough examination due to their substantial training complexities. To address these gaps, this paper extends the original environment by not only standardizing the environment settings and benchmarking cooperative learning algorithms across different scenarios, including the most challenging full-game scenarios, but also by discussing approaches to enhance football AI from diverse perspectives and introducing related research tools. Specifically, we provide a distributed and asynchronous population-based self-play framework with diverse pre-trained policies for faster training, two football-specific analytical tools for deeper investigation, and an online leaderboard for broader evaluation. The overall expectation of this work is to advance the study of Multi-Agent Reinforcement Learning on Google Research Football environment, with the ultimate goal of benefiting real-world sports beyond virtual games.

Offline multi-agent reinforcement learning is challenging due to the coupling effect of both distribution shift issue common in offline setting and the high dimension issue common in multi-agent setting, making the action out-of-distribution (OOD) and value overestimation phenomenon excessively severe. To mitigate this problem, we propose a novel multi-agent offline RL algorithm, named CounterFactual Conservative Q-Learning (CFCQL) to conduct conservative value estimation. Rather than regarding all the agents as a high dimensional single one and directly applying single agent methods to it, CFCQL calculates conservative regularization for each agent separately in a counterfactual way and then linearly combines them to realize an overall conservative value estimation. We prove that it still enjoys the underestimation property and the performance guarantee as those single agent conservative methods do, but the induced regularization and safe policy improvement bound are independent of the agent number, which is therefore theoretically superior to the direct treatment referred to above, especially when the agent number is large. We further conduct experiments on four environments including both discrete and continuous action settings on both existing and our man-made datasets, demonstrating that CFCQL outperforms existing methods on most datasets and even with a remarkable margin on some of them.

In open multi-agent environments, the agents may encounter unexpected teammates. Classical multi-agent learning approaches train agents that can only coordinate with seen teammates. Recent studies attempted to generate diverse teammates to enhance the generalizable coordination ability, but were restricted by pre-defined teammates. In this work, our aim is to train agents with strong coordination ability by generating teammates that fully cover the teammate policy space, so that agents can coordinate with any teammates. Since the teammate policy space is too huge to be enumerated, we find only dissimilar teammates that are incompatible with controllable agents, which highly reduces the number of teammates that need to be trained with. However, it is hard to determine the number of such incompatible teammates beforehand. We therefore introduce a continual multi-agent learning process, in which the agent learns to coordinate with different teammates until no more incompatible teammates can be found. The above idea is implemented in the proposed Macop (Multi-agent compatible policy learning) algorithm. We conduct experiments in 8 scenarios from 4 environments that have distinct coordination patterns. Experiments show that Macop generates training teammates with much lower compatibility than previous methods. As a result, in all scenarios Macop achieves the best overall coordination ability while never significantly worse than the baselines, showing strong generalization ability.

Being able to harness the power of large datasets for developing cooperative multi-agent controllers promises to unlock enormous value for real-world applications. Many important industrial systems are multi-agent in nature and are difficult to model using bespoke simulators. However, in industry, distributed processes can often be recorded during operation, and large quantities of demonstrative data stored. Offline multi-agent reinforcement learning (MARL) provides a promising paradigm for building effective decentralised controllers from such datasets. However, offline MARL is still in its infancy and therefore lacks standardised benchmark datasets and baselines typically found in more mature subfields of reinforcement learning (RL). These deficiencies make it difficult for the community to sensibly measure progress. In this work, we aim to fill this gap by releasing off-the-grid MARL (OG-MARL): a growing repository of high-quality datasets with baselines for cooperative offline MARL research. Our datasets provide settings that are characteristic of real-world systems, including complex environment dynamics, heterogeneous agents, non-stationarity, many agents, partial observability, suboptimality, sparse rewards and demonstrated coordination. For each setting, we provide a range of different dataset types (e.g. Good, Medium, Poor, and Replay) and profile the composition of experiences for each dataset. We hope that OG-MARL will serve the community as a reliable source of datasets and help drive progress, while also providing an accessible entry point for researchers new to the field.

Conventional text style transfer approaches focus on sentence-level style transfer without considering contextual information, and the style is described with attributes (e.g., formality). When applying style transfer in conversations such as task-oriented dialogues, existing approaches suffer from these limitations as context can play an important role and the style attributes are often difficult to define in conversations. In this paper, we introduce conversation style transfer as a few-shot learning problem, where the model learns to perform style transfer by observing only a few example dialogues in the target style. We propose a novel in-context learning approach to solve the task with style-free dialogues as a pivot. Human evaluation shows that by incorporating multi-turn context, the model is able to match the target style while having better appropriateness and semantic correctness compared to utterance/sentence-level style transfer. Additionally, we show that conversation style transfer can also benefit downstream tasks. For example, in multi-domain intent classification tasks, the F1 scores improve after transferring the style of training data to match the style of the test data.

Distributed optimization is a fundamental framework for collaborative inference and decision making in decentralized multi-agent systems. The operation is modeled as the joint minimization of a shared objective which typically depends on observations gathered locally by each agent. Distributed optimization algorithms, such as the common D-ADMM, tackle this task by iteratively combining local computations and message exchanges. One of the main challenges associated with distributed optimization, and particularly with D-ADMM, is that it requires a large number of communications, i.e., messages exchanged between the agents, to reach consensus. This can make D-ADMM costly in power, latency, and channel resources. In this work we propose unfolded D-ADMM, which follows the emerging deep unfolding methodology to enable D-ADMM to operate reliably with a predefined and small number of messages exchanged by each agent. Unfolded D-ADMM fully preserves the operation of D-ADMM, while leveraging data to tune the hyperparameters of each iteration of the algorithm. These hyperparameters can either be agent-specific, aiming at achieving the best performance within a fixed number of iterations over a given network, or shared among the agents, allowing to learn to distributedly optimize over different networks. For both settings, our unfolded D-ADMM operates with limited communications, while preserving the interpretability and flexibility of the original D-ADMM algorithm. We specialize unfolded D-ADMM for two representative settings: a distributed estimation task, considering a sparse recovery setup, and a distributed learning scenario, where multiple agents collaborate in learning a machine learning model. Our numerical results demonstrate that the proposed approach dramatically reduces the number of communications utilized by D-ADMM, without compromising on its performance.

We present a distributed conjugate gradient method for distributed optimization problems, where each agent computes an optimal solution of the problem locally without any central computation or coordination, while communicating with its immediate, one-hop neighbors over a communication network. Each agent updates its local problem variable using an estimate of the average conjugate direction across the network, computed via a dynamic consensus approach. Our algorithm enables the agents to use uncoordinated step-sizes. We prove convergence of the local variable of each agent to the optimal solution of the aggregate optimization problem, without requiring decreasing step-sizes. In addition, we demonstrate the efficacy of our algorithm in distributed state estimation problems, and its robust counterparts, where we show its performance compared to existing distributed first-order optimization methods.

A fundamental shortcoming of the concept of Nash equilibrium is its computational intractability: approximating Nash equilibria in normal-form games is PPAD-hard. In this paper, inspired by the ideas of smoothed analysis, we introduce a relaxed variant of Nash equilibrium called $\sigma$-smooth Nash equilibrium, for a smoothness parameter $\sigma$. In a $\sigma$-smooth Nash equilibrium, players only need to achieve utility at least as high as their best deviation to a $\sigma$-smooth strategy, which is a distribution that does not put too much mass (as parametrized by $\sigma$) on any fixed action. We distinguish two variants of $\sigma$-smooth Nash equilibria: strong $\sigma$-smooth Nash equilibria, in which players are required to play $\sigma$-smooth strategies under equilibrium play, and weak $\sigma$-smooth Nash equilibria, where there is no such requirement. We show that both weak and strong $\sigma$-smooth Nash equilibria have superior computational properties to Nash equilibria: when $\sigma$ as well as an approximation parameter $\epsilon$ and the number of players are all constants, there is a constant-time randomized algorithm to find a weak $\epsilon$-approximate $\sigma$-smooth Nash equilibrium in normal-form games. In the same parameter regime, there is a polynomial-time deterministic algorithm to find a strong $\epsilon$-approximate $\sigma$-smooth Nash equilibrium in a normal-form game. These results stand in contrast to the optimal algorithm for computing $\epsilon$-approximate Nash equilibria, which cannot run in faster than quasipolynomial-time. We complement our upper bounds by showing that when either $\sigma$ or $\epsilon$ is an inverse polynomial, finding a weak $\epsilon$-approximate $\sigma$-smooth Nash equilibria becomes computationally intractable.

As machine learning models become more general, we need to characterise them in richer, more meaningful ways. We describe a method to infer the cognitive profile of a system from diverse experimental data. To do so, we introduce measurement layouts that model how task-instance features interact with system capabilities to affect performance. These features must be triangulated in complex ways to be able to infer capabilities from non-populational data -- a challenge for traditional psychometric and inferential tools. Using the Bayesian probabilistic programming library PyMC, we infer different cognitive profiles for agents in two scenarios: 68 actual contestants in the AnimalAI Olympics and 30 synthetic agents for O-PIAAGETS, an object permanence battery. We showcase the potential for capability-oriented evaluation.