Weak-to-Strong Preference Optimization: Stealing Reward from Weak Aligned Model

• URL: http://arxiv.org/abs/2410.18640v1
• Date: Thu, 24 Oct 2024 11:06:29 GMT
• Title: Weak-to-Strong Preference Optimization: Stealing Reward from Weak Aligned Model
• Authors: Wenhong Zhu, Zhiwei He, Xiaofeng Wang, Pengfei Liu, Rui Wang,
• Abstract summary: This work is inspired by weak-to-strong generalization, where a strong LM fine-tuned on labels generated by a weaker model can consistently outperform its weak supervisor. We propose Weak-to-Strong Preference Optimization (WSPO), which achieves strong model alignment by learning the distribution differences before and after the alignment of the weak model.
• Score: 28.569089876442682
• License:
• Abstract: Aligning language models (LMs) with human preferences has become a key area of research, enabling these models to meet diverse user needs better. Inspired by weak-to-strong generalization, where a strong LM fine-tuned on labels generated by a weaker model can consistently outperform its weak supervisor, we extend this idea to model alignment. In this work, we observe that the alignment behavior in weaker models can be effectively transferred to stronger models and even exhibit an amplification effect. Based on this insight, we propose a method called Weak-to-Strong Preference Optimization (WSPO), which achieves strong model alignment by learning the distribution differences before and after the alignment of the weak model. Experiments demonstrate that WSPO delivers outstanding performance, improving the win rate of Qwen2-7B-Instruct on Arena-Hard from 39.70 to 49.60, achieving a remarkable 47.04 length-controlled win rate on AlpacaEval 2, and scoring 7.33 on MT-bench. Our results suggest that using the weak model to elicit a strong model with a high alignment ability is feasible.

Related papers

• RM-Bench: Benchmarking Reward Models of Language Models with Subtlety and Style [37.97757796124621]
  RM-Bench is a novel benchmark designed to evaluate reward models based on their sensitivity to subtle content differences and resistance to style biases. We evaluate nearly 40 reward models on RM-Bench and find that even state-of-the-art models achieve an average performance of only 46.6%.
  arXiv  Detail & Related papers  (2024-10-21T16:48:26Z)
• Margin Matching Preference Optimization: Enhanced Model Alignment with Granular Feedback [64.67540769692074]
  Large language models (LLMs) fine-tuned with alignment techniques, such as reinforcement learning from human feedback, have been instrumental in developing some of the most capable AI systems to date. We introduce an approach called Margin Matching Preference Optimization (MMPO), which incorporates relative quality margins into optimization, leading to improved LLM policies and reward models. Experiments with both human and AI feedback data demonstrate that MMPO consistently outperforms baseline methods, often by a substantial margin, on popular benchmarks including MT-bench and RewardBench.
  arXiv  Detail & Related papers  (2024-10-04T04:56:11Z)
• Self-Taught Evaluators [77.92610887220594]
  We present an approach that aims to im-proves without human annotations, using synthetic training data only. Our Self-Taught Evaluator can improve a strong LLM from 75.4 to 88.3 on RewardBench.
  arXiv  Detail & Related papers  (2024-08-05T17:57:02Z)
• Weak-to-Strong Reasoning [33.20094938292376]
  We introduce a progressive learning framework that enables the strong model to autonomously refine its training data. Our method significantly enhances the reasoning capabilities of Llama2-70b using three separate weak models. This work paves the way for a more scalable and sophisticated strategy to enhance AI reasoning powers.
  arXiv  Detail & Related papers  (2024-07-18T16:25:17Z)
• Super(ficial)-alignment: Strong Models May Deceive Weak Models in Weak-to-Strong Generalization [68.62228569439478]
  We investigate whether there exists an issue of weak-to-strong deception. We find that the deception intensifies as the capability gap between weak and strong models increases. Our work highlights the urgent need to pay more attention to the true reliability of superalignment.
  arXiv  Detail & Related papers  (2024-06-17T11:36:39Z)
• Quantifying the Gain in Weak-to-Strong Generalization [14.453654853392619]
  We show that the improvement in performance achieved by strong models over their weaker counterparts is quantified by the misfit error incurred by the strong model on labels generated by the weaker model. For instance, we can predict the amount by which the strong model will improve over the weak model, and also choose among different weak models to train the strong model, based on its misfit error.
  arXiv  Detail & Related papers  (2024-05-24T00:14:16Z)
• Self-Play Preference Optimization for Language Model Alignment [75.83359213697854]
  Recent advancements suggest that directly working with preference probabilities can yield a more accurate reflection of human preferences. We propose a self-play-based method for language model alignment, which treats the problem as a constant-sum two-player game. Our approach, dubbed Self-Play Preference Optimization (SPPO), utilizes iterative policy updates to provably approximate the Nash equilibrium.
  arXiv  Detail & Related papers  (2024-05-01T17:59:20Z)
• Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision [55.196139002977525]
  Superhuman models will behave in complex ways too difficult for humans to reliably evaluate. Can weak model supervision elicit the full capabilities of a much stronger model? We find that when we naively finetune strong pretrained models on labels generated by a weak model, they consistently perform better than their weak supervisors.
  arXiv  Detail & Related papers  (2023-12-14T23:07:33Z)
• Voting based ensemble improves robustness of defensive models [82.70303474487105]
  We study whether it is possible to create an ensemble to further improve robustness. By ensembling several state-of-the-art pre-trained defense models, our method can achieve a 59.8% robust accuracy.
  arXiv  Detail & Related papers  (2020-11-28T00:08:45Z)
• Adversarial Concurrent Training: Optimizing Robustness and Accuracy Trade-off of Deep Neural Networks [13.041607703862724]
  We propose Adversarial Concurrent Training (ACT) to train a robust model in conjunction with a natural model in a minimax game. ACT achieves 68.20% standard accuracy and 44.29% robustness accuracy under a 100-iteration untargeted attack.
  arXiv  Detail & Related papers  (2020-08-16T22:14:48Z)

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.