Magnetic Preference Optimization: Achieving Last-iterate Convergence for Language Models Alignment

Self-play methods have demonstrated remarkable success in enhancing model capabilities across various domains. In the context of Reinforcement Learning from Human Feedback (RLHF), self-play not only boosts Large Language Model (LLM) performance but also overcomes the limitations of traditional Bradley-Terry (BT) model assumptions by finding the Nash equilibrium (NE) of a preference-based, two-player constant-sum game. However, existing methods either guarantee only average-iterate convergence, incurring high storage and inference costs, or converge to the NE of a regularized game, failing to accurately reflect true human preferences. In this paper, we introduce Magnetic Preference Optimization (MPO), a novel approach capable of achieving last-iterate convergence to the NE of the original game, effectively overcoming the limitations of existing methods. Building upon Magnetic Mirror Descent (MMD), MPO attains a linear convergence rate, making it particularly suitable for fine-tuning LLMs. To ensure our algorithm is both theoretically sound and practically viable, we present a simple yet effective implementation that adapts the theoretical insights to the RLHF setting. Empirical results demonstrate that MPO can significantly enhance the performance of LLMs, highlighting the potential of self-play methods in alignment.

A Survey on Self-play Methods in Reinforcement Learning

Self-play, characterized by agents' interactions with copies or past versions of itself, has recently gained prominence in reinforcement learning. This paper first clarifies the preliminaries of self-play, including the multi-agent reinforcement learning framework and basic game theory concepts. Then it provides a unified framework and classifies existing self-play algorithms within this framework. Moreover, the paper bridges the gap between the algorithms and their practical implications by illustrating the role of self-play in different scenarios. Finally, the survey highlights open challenges and future research directions in self-play. This paper is an essential guide map for understanding the multifaceted landscape of self-play in RL.

Fusion-PSRO: Nash Policy Fusion for Policy Space Response Oracles

A popular approach for solving zero-sum games is to maintain populations of policies to approximate the Nash Equilibrium (NE). Previous studies have shown that Policy Space Response Oracle (PSRO) algorithm is an effective multi-agent reinforcement learning framework for solving such games. However, repeatedly training new policies from scratch to approximate Best Response (BR) to opponents' mixed policies at each iteration is both inefficient and costly. While some PSRO variants initialize a new policy by inheriting from past BR policies, this approach limits the exploration of new policies, especially against challenging opponents. To address this issue, we propose Fusion-PSRO, which employs policy fusion to initialize policies for better approximation to BR. By selecting high-quality base policies from meta-NE, policy fusion fuses the base policies into a new policy through model averaging. This approach allows the initialized policies to incorporate multiple expert policies, making it easier to handle difficult opponents compared to inheriting from past BR policies or initializing from scratch. Moreover, our method only modifies the policy initialization phase, allowing its application to nearly all PSRO variants without additional training overhead. Our experiments on non-transitive matrix games, Leduc Poker, and the more complex Liars Dice demonstrate that Fusion-PSRO enhances the performance of nearly all PSRO variants, achieving lower exploitability.

Incentive Compatibility for AI Alignment in Sociotechnical Systems: Positions and Prospects

The burgeoning integration of artificial intelligence (AI) into human society brings forth significant implications for societal governance and safety. While considerable strides have been made in addressing AI alignment challenges, existing methodologies primarily focus on technical facets, often neglecting the intricate sociotechnical nature of AI systems, which can lead to a misalignment between the development and deployment contexts. To this end, we posit a new problem worth exploring: Incentive Compatibility Sociotechnical Alignment Problem (ICSAP). We hope this can call for more researchers to explore how to leverage the principles of Incentive Compatibility (IC) from game theory to bridge the gap between technical and societal components to maintain AI consensus with human societies in different contexts. We further discuss three classical game problems for achieving IC: mechanism design, contract theory, and Bayesian persuasion, in addressing the perspectives, potentials, and challenges of solving ICSAP, and provide preliminary implementation conceptions.

Panacea: Pareto Alignment via Preference Adaptation for LLMs

Current methods for large language model alignment typically use scalar human preference labels. However, this convention tends to oversimplify the multi-dimensional and heterogeneous nature of human preferences, leading to reduced expressivity and even misalignment. This paper presents Panacea, an innovative approach that reframes alignment as a multi-dimensional preference optimization problem. Panacea trains a single model capable of adapting online and Pareto-optimally to diverse sets of preferences without the need for further tuning. A major challenge here is using a low-dimensional preference vector to guide the model's behavior, despite it being governed by an overwhelmingly large number of parameters. To address this, Panacea is designed to use singular value decomposition (SVD)-based low-rank adaptation, which allows the preference vector to be simply injected online as singular values. Theoretically, we prove that Panacea recovers the entire Pareto front with common loss aggregation methods under mild conditions. Moreover, our experiments demonstrate, for the first time, the feasibility of aligning a single LLM to represent a spectrum of human preferences through various optimization methods. Our work marks a step forward in effectively and efficiently aligning models to diverse and intricate human preferences in a controllable and Pareto-optimal manner.

Red Teaming Game: A Game-Theoretic Framework for Red Teaming Language Models

Deployable Large Language Models (LLMs) must conform to the criterion of helpfulness and harmlessness, thereby achieving consistency between LLMs outputs and human values. Red-teaming techniques constitute a critical way towards this criterion. Existing work rely solely on manual red team designs and heuristic adversarial prompts for vulnerability detection and optimization. These approaches lack rigorous mathematical formulation, thus limiting the exploration of diverse attack strategy within quantifiable measure and optimization of LLMs under convergence guarantees. In this paper, we present Red-teaming Game (RTG), a general game-theoretic framework without manual annotation. RTG is designed for analyzing the multi-turn attack and defense interactions between Red-team language Models (RLMs) and Blue-team Language Model (BLM). Within the RTG, we propose Gamified Red-teaming Solver (GRTS) with diversity measure of the semantic space. GRTS is an automated red teaming technique to solve RTG towards Nash equilibrium through meta-game analysis, which corresponds to the theoretically guaranteed optimization direction of both RLMs and BLM. Empirical results in multi-turn attacks with RLMs show that GRTS autonomously discovered diverse attack strategies and effectively improved security of LLMs, outperforming existing heuristic red-team designs. Overall, RTG has established a foundational framework for red teaming tasks and constructed a new scalable oversight technique for alignment.

Fully Decentralized Model-based Policy Optimization for Networked Systems

Reinforcement learning algorithms require a large amount of samples; this often limits their real-world applications on even simple tasks. Such a challenge is more outstanding in multi-agent tasks, as each step of operation is more costly requiring communications or shifting or resources. This work aims to improve data efficiency of multi-agent control by model-based learning. We consider networked systems where agents are cooperative and communicate only locally with their neighbors, and propose the decentralized model-based policy optimization framework (DMPO). In our method, each agent learns a dynamic model to predict future states and broadcast their predictions by communication, and then the policies are trained under the model rollouts. To alleviate the bias of model-generated data, we restrain the model usage for generating myopic rollouts, thus reducing the compounding error of model generation. To pertain the independence of policy update, we introduce extended value function and theoretically prove that the resulting policy gradient is a close approximation to true policy gradients. We evaluate our algorithm on several benchmarks for intelligent transportation systems, which are connected autonomous vehicle control tasks (Flow and CACC) and adaptive traffic signal control (ATSC). Empirically results show that our method achieves superior data efficiency and matches the performance of model-free methods using true models.