COMAL: A Convergent Meta-Algorithm for Aligning LLMs with General Preferences

• URL: http://arxiv.org/abs/2410.23223v1
• Date: Wed, 30 Oct 2024 17:13:02 GMT
• Title: COMAL: A Convergent Meta-Algorithm for Aligning LLMs with General Preferences
• Authors: Yixin Liu, Argyris Oikonomou, Weiqiang Zheng, Yang Cai, Arman Cohan,
• Abstract summary: We propose a meta-algorithm, Convergent Meta Alignment Algorithm (COMAL), for language model alignment with general preferences. Our meta-algorithm is simple and can be integrated with many existing methods designed for RLHF and preference optimization.
• Score: 31.988100672680154
• License:
• Abstract: Many alignment methods, including reinforcement learning from human feedback (RLHF), rely on the Bradley-Terry reward assumption, which is insufficient to capture the full range of general human preferences. To achieve robust alignment with general preferences, we model the alignment problem as a two-player zero-sum game, where the Nash equilibrium policy guarantees a 50% win rate against any competing policy. However, previous algorithms for finding the Nash policy either diverge or converge to a Nash policy in a modified game, even in a simple synthetic setting, thereby failing to maintain the 50% win rate guarantee against all other policies. We propose a meta-algorithm, Convergent Meta Alignment Algorithm (COMAL), for language model alignment with general preferences, inspired by convergent algorithms in game theory. Theoretically, we prove that our meta-algorithm converges to an exact Nash policy in the last iterate. Additionally, our meta-algorithm is simple and can be integrated with many existing methods designed for RLHF and preference optimization with minimal changes. Experimental results demonstrate the effectiveness of the proposed framework when combined with existing preference policy optimization methods.

Related papers

• Iterative Nash Policy Optimization: Aligning LLMs with General Preferences via No-Regret Learning [55.65738319966385]
  We propose a novel online algorithm, iterative Nash policy optimization (INPO) Unlike previous methods, INPO bypasses the need for estimating the expected win rate for individual responses. With an LLaMA-3-8B-based SFT model, INPO achieves a 42.6% length-controlled win rate on AlpacaEval 2.0 and a 37.8% win rate on Arena-Hard.
  arXiv  Detail & Related papers  (2024-06-30T08:00:34Z)
• e-COP : Episodic Constrained Optimization of Policies [12.854752753529151]
  We present the first policy optimization algorithm for constrained Reinforcement Learning (RL) in episodic (finite horizon) settings. We show that our algorithm has similar or better performance than SoTA (non-episodic) algorithms adapted for the episodic setting.
  arXiv  Detail & Related papers  (2024-06-13T20:12:09Z)
• Bridging the Gap between Newton-Raphson Method and Regularized Policy Iteration [13.166738075816493]
  We show that regularized policy iteration is strictly equivalent to the standard Newton-Raphson method in the condition of smoothing out Bellman equation with strongly convex functions. We prove that regularized policy iteration has global linear convergence with the rate being $gamma$ (discount factor) We also show that a modified version of regularized policy iteration, i.e., with finite-step policy evaluation, is equivalent to inexact Newton method where the Newton iteration formula is solved with truncated iterations.
  arXiv  Detail & Related papers  (2023-10-11T05:55:20Z)
• Low-Switching Policy Gradient with Exploration via Online Sensitivity Sampling [23.989009116398208]
  We design a low-switching sample-efficient policy optimization algorithm, LPO, with general non-linear function approximation. We show that, our algorithm obtains an $varepsilon$-optimal policy with only $widetildeO(fractextpoly(d)varepsilon3)$ samples.
  arXiv  Detail & Related papers  (2023-06-15T23:51:46Z)
• A New Policy Iteration Algorithm For Reinforcement Learning in Zero-Sum Markov Games [10.805520579293747]
  We show that a simple variant of naive policy iteration for games converges exponentially fast. We also show that lookahead policies can be implemented efficiently in the function approximation setting of linear Markov games.
  arXiv  Detail & Related papers  (2023-03-17T01:20:22Z)
• Maximum-Likelihood Inverse Reinforcement Learning with Finite-Time Guarantees [56.848265937921354]
  Inverse reinforcement learning (IRL) aims to recover the reward function and the associated optimal policy. Many algorithms for IRL have an inherently nested structure. We develop a novel single-loop algorithm for IRL that does not compromise reward estimation accuracy.
  arXiv  Detail & Related papers  (2022-10-04T17:13:45Z)
• Faster Last-iterate Convergence of Policy Optimization in Zero-Sum Markov Games [63.60117916422867]
  This paper focuses on the most basic setting of competitive multi-agent RL, namely two-player zero-sum Markov games. We propose a single-loop policy optimization method with symmetric updates from both agents, where the policy is updated via the entropy-regularized optimistic multiplicative weights update (OMWU) method. Our convergence results improve upon the best known complexities, and lead to a better understanding of policy optimization in competitive Markov games.
  arXiv  Detail & Related papers  (2022-10-03T16:05:43Z)
• Policy Optimization for Markov Games: Unified Framework and Faster Convergence [81.3266426402464]
  We show that the state-wise average policy of this algorithm converges to an approximate Nash equilibrium (NE) of the game. We extend this algorithm to multi-player general Markov Games and show a $mathcalwidetildeO(T-1/2)$ convergence rate to Correlated Equilibria (CCE)
  arXiv  Detail & Related papers  (2022-06-06T14:23:13Z)
• Understanding the Effect of Stochasticity in Policy Optimization [86.7574122154668]
  We show that the preferability of optimization methods depends critically on whether exact gradients are used. Second, to explain these findings we introduce the concept of committal rate for policy optimization. Third, we show that in the absence of external oracle information, there is an inherent trade-off between exploiting geometry to accelerate convergence versus achieving optimality almost surely.
  arXiv  Detail & Related papers  (2021-10-29T06:35:44Z)
• A Policy Efficient Reduction Approach to Convex Constrained Deep Reinforcement Learning [2.811714058940267]
  We propose a new variant of the conditional gradient (CG) type algorithm, which generalizes the minimum norm point (MNP) method. Our method reduces the memory costs by an order of magnitude, and achieves better performance, demonstrating both its effectiveness and efficiency.
  arXiv  Detail & Related papers  (2021-08-29T20:51:32Z)
• Provable Fictitious Play for General Mean-Field Games [111.44976345867005]
  We propose a reinforcement learning algorithm for stationary mean-field games. The goal is to learn a pair of mean-field state and stationary policy that constitutes the Nash equilibrium.
  arXiv  Detail & Related papers  (2020-10-08T18:46:48Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.