Learn Your Reference Model for Real Good Alignment

• URL: http://arxiv.org/abs/2404.09656v3
• Date: Fri, 11 Oct 2024 13:42:12 GMT
• Title: Learn Your Reference Model for Real Good Alignment
• Authors: Alexey Gorbatovski, Boris Shaposhnikov, Alexey Malakhov, Nikita Surnachev, Yaroslav Aksenov, Ian Maksimov, Nikita Balagansky, Daniil Gavrilov,
• Abstract summary: offline methods for Large Language Models (LLMs) alignment are susceptible to overoptimization. We propose a new paradigm of offline alignment methods, called Trust Region, which dynamically updates the reference policy throughout the training process. Our results show that TR alignment methods effectively mitigate overoptimization, enabling models to maintain strong performance even when substantially deviating from the initial reference policy.
• Score: 3.091688550418396
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Despite the fact that offline methods for Large Language Models (LLMs) alignment do not require a direct reward model, they remain susceptible to overoptimization. This issue arises when the trained model deviates excessively from the reference policy, leading to a decrease in sample quality. We propose a new paradigm of offline alignment methods, called Trust Region (including variants TR-DPO, TR-IPO, TR-KTO), which dynamically updates the reference policy throughout the training process. Our results show that TR alignment methods effectively mitigate overoptimization, enabling models to maintain strong performance even when substantially deviating from the initial reference policy. We demonstrate the efficacy of these approaches not only through toy examples that exhibit reduced overoptimization, but also through direct, side-by-side comparisons in specific tasks such as helpful and harmless dialogue, as well as summarization, where they surpass conventional methods. Additionally, we report significant improvements in general-purpose assistant setups with the Llama3 model on the AlpacaEval 2 and Arena-Hard benchmarks, highlighting the advantages of Trust Region methods over classical approaches.

Related papers

• SePPO: Semi-Policy Preference Optimization for Diffusion Alignment [67.8738082040299]
  We propose a preference optimization method that aligns DMs with preferences without relying on reward models or paired human-annotated data. We validate SePPO across both text-to-image and text-to-video benchmarks.
  arXiv  Detail & Related papers  (2024-10-07T17:56:53Z)
• Correcting the Mythos of KL-Regularization: Direct Alignment without Overoptimization via Chi-Squared Preference Optimization [78.82586283794886]
  $chi2$-Preference Optimization ($chi$PO) is an efficient offline alignment algorithm provably robust to overoptimization. $chi$PO implements the principle of pessimism in the face of uncertainty via regularization. $chi$PO's simplicity and strong guarantees make it the first practical and general-purpose offline alignment algorithm provably robust to overoptimization.
  arXiv  Detail & Related papers  (2024-07-18T11:08:40Z)
• Contrastive Policy Gradient: Aligning LLMs on sequence-level scores in a supervised-friendly fashion [44.95386817008473]
  We introduce Contrastive Policy Gradient, or CoPG, a simple and mathematically principled new RL algorithm that can estimate the optimal policy even from off-policy data. We show this approach to generalize the direct alignment method IPO (identity preference optimization) and classic policy gradient. We experiment with the proposed CoPG on a toy bandit problem to illustrate its properties, as well as for finetuning LLMs on a summarization task.
  arXiv  Detail & Related papers  (2024-06-27T14:03:49Z)
• SAIL: Self-Improving Efficient Online Alignment of Large Language Models [56.59644677997827]
  Reinforcement Learning from Human Feedback is a key method for aligning large language models with human preferences. Recent literature has focused on designing online RLHF methods but still lacks a unified conceptual formulation. Our approach significantly improves alignment performance on open-sourced datasets with minimal computational overhead.
  arXiv  Detail & Related papers  (2024-06-21T18:05:35Z)
• Self-Play with Adversarial Critic: Provable and Scalable Offline Alignment for Language Models [44.38073745307387]
  This work studies the challenge of aligning large language models (LLMs) with offline preference data. We propose SPAC, a new offline preference optimization method with self-play, inspired by the on-average pessimism technique from the offline RL literature.
  arXiv  Detail & Related papers  (2024-06-06T17:23:49Z)
• Self-Augmented Preference Optimization: Off-Policy Paradigms for Language Model Alignment [104.18002641195442]
  We introduce Self-Augmented Preference Optimization (SAPO), an effective and scalable training paradigm that does not require existing paired data. Building on the self-play concept, which autonomously generates negative responses, we further incorporate an off-policy learning pipeline to enhance data exploration and exploitation.
  arXiv  Detail & Related papers  (2024-05-31T14:21:04Z)
• Optimal Baseline Corrections for Off-Policy Contextual Bandits [61.740094604552475]
  We aim to learn decision policies that optimize an unbiased offline estimate of an online reward metric. We propose a single framework built on their equivalence in learning scenarios. Our framework enables us to characterize the variance-optimal unbiased estimator and provide a closed-form solution for it.
  arXiv  Detail & Related papers  (2024-05-09T12:52:22Z)
• Model-based trajectory stitching for improved behavioural cloning and its applications [7.462336024223669]
  Trajectory Stitching (TS) generates new trajectories by stitching' pairs of states that were disconnected in the original data. We demonstrate that the iterative process of replacing old trajectories with new ones incrementally improves the underlying behavioural policy.
  arXiv  Detail & Related papers  (2022-12-08T14:18:04Z)
• Supported Policy Optimization for Offline Reinforcement Learning [74.1011309005488]
  Policy constraint methods to offline reinforcement learning (RL) typically utilize parameterization or regularization. Regularization methods reduce the divergence between the learned policy and the behavior policy. This paper presents Supported Policy OpTimization (SPOT), which is directly derived from the theoretical formalization of the density-based support constraint.
  arXiv  Detail & Related papers  (2022-02-13T07:38:36Z)
• COMBO: Conservative Offline Model-Based Policy Optimization [120.55713363569845]
  Uncertainty estimation with complex models, such as deep neural networks, can be difficult and unreliable. We develop a new model-based offline RL algorithm, COMBO, that regularizes the value function on out-of-support state-actions. We find that COMBO consistently performs as well or better as compared to prior offline model-free and model-based methods.
  arXiv  Detail & Related papers  (2021-02-16T18:50:32Z)

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.