Iterative Label Refinement Matters More than Preference Optimization under Weak Supervision

• URL: http://arxiv.org/abs/2501.07886v1
• Date: Tue, 14 Jan 2025 06:54:17 GMT
• Title: Iterative Label Refinement Matters More than Preference Optimization under Weak Supervision
• Authors: Yaowen Ye, Cassidy Laidlaw, Jacob Steinhardt,
• Abstract summary: We simulate unreliable demonstrations and comparison feedback using small language models and humans. We find that in the presence of unreliable supervision, SFT still retains some effectiveness, but DPO (a common RLHF algorithm) fails to improve the model beyond SFT. Our findings suggest that as LMs are used for complex tasks where human supervision is unreliable, RLHF may no longer be the best use of human comparison feedback.
• Score: 34.594109869213014
• License:
• Abstract: Language model (LM) post-training relies on two stages of human supervision: task demonstrations for supervised finetuning (SFT), followed by preference comparisons for reinforcement learning from human feedback (RLHF). As LMs become more capable, the tasks they are given become harder to supervise. Will post-training remain effective under unreliable supervision? To test this, we simulate unreliable demonstrations and comparison feedback using small LMs and time-constrained humans. We find that in the presence of unreliable supervision, SFT still retains some effectiveness, but DPO (a common RLHF algorithm) fails to improve the model beyond SFT. To address this, we propose iterative label refinement (ILR) as an alternative to RLHF. ILR improves the SFT data by using comparison feedback to decide whether human demonstrations should be replaced by model-generated alternatives, then retrains the model via SFT on the updated data. SFT+ILR outperforms SFT+DPO on several tasks with unreliable supervision (math, coding, and safe instruction-following). Our findings suggest that as LMs are used for complex tasks where human supervision is unreliable, RLHF may no longer be the best use of human comparison feedback; instead, it is better to direct feedback towards improving the training data rather than continually training the model. Our code and data are available at https://github.com/helloelwin/iterative-label-refinement.

Related papers

• ASFT: Aligned Supervised Fine-Tuning through Absolute Likelihood [14.512464277772194]
  Aligned Supervised Fine-Tuning (ASFT) is an effective approach that better aligns Large Language Models with pair-wise datasets. ASFT mitigates the issue where the DPO loss function decreases the probability of generating human-dispreferred data. Extensive experiments demonstrate that ASFT is an effective alignment approach, consistently outperforming existing methods.
  arXiv  Detail & Related papers  (2024-09-14T11:39:13Z)
• Unlock the Correlation between Supervised Fine-Tuning and Reinforcement Learning in Training Code Large Language Models [12.656574142412484]
  We make an attempt to understand the correlation between supervised fine-tuning and reinforcement learning. We find that both atomic and synthetic functions are indispensable for SFT's generalization.
  arXiv  Detail & Related papers  (2024-06-14T03:39:01Z)
• Getting More Juice Out of the SFT Data: Reward Learning from Human Demonstration Improves SFT for LLM Alignment [65.15914284008973]
  We propose to leverage an Inverse Reinforcement Learning (IRL) technique to simultaneously build an reward model and a policy model. We show that the proposed algorithms converge to the stationary solutions of the IRL problem. Our results indicate that it is beneficial to leverage reward learning throughout the entire alignment process.
  arXiv  Detail & Related papers  (2024-05-28T07:11:05Z)
• A Critical Evaluation of AI Feedback for Aligning Large Language Models [60.42291111149438]
  We show that simple supervised fine-tuning with GPT-4 as the teacher outperforms existing RLAIF pipelines. More generally, we find that the gains from RLAIF vary substantially across base model families, test-time evaluation protocols, and critic models.
  arXiv  Detail & Related papers  (2024-02-19T18:53:54Z)
• SteerLM: Attribute Conditioned SFT as an (User-Steerable) Alternative to RLHF [19.43122743768123]
  We propose SteerLM, a supervised fine-tuning method that empowers end-users to control responses during inference. SteerLM conditions responses to conform to an explicitly defined multi-dimensional set of attributes, thereby empowering a steerable AI capable of generating helpful and high-quality responses.
  arXiv  Detail & Related papers  (2023-10-09T02:11:21Z)
• Scaling Relationship on Learning Mathematical Reasoning with Large Language Models [75.29595679428105]
  We investigate how the pre-training loss, supervised data amount, and augmented data amount influence the reasoning performances of a supervised LLM. We find that rejection samples from multiple models push LLaMA-7B to an accuracy of 49.3% on GSM8K which outperforms the supervised fine-tuning (SFT) accuracy of 35.9% significantly.
  arXiv  Detail & Related papers  (2023-08-03T15:34:01Z)
• Value function estimation using conditional diffusion models for control [62.27184818047923]
  We propose a simple algorithm called Diffused Value Function (DVF) It learns a joint multi-step model of the environment-robot interaction dynamics using a diffusion model. We show how DVF can be used to efficiently capture the state visitation measure for multiple controllers.
  arXiv  Detail & Related papers  (2023-06-09T18:40:55Z)
• Direct Preference Optimization: Your Language Model is Secretly a Reward Model [119.65409513119963]
  We introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods.
  arXiv  Detail & Related papers  (2023-05-29T17:57:46Z)
• Improving Representational Continuity via Continued Pretraining [76.29171039601948]
  Transfer learning community (LP-FT) outperforms naive training and other continual learning methods. LP-FT also reduces forgetting in a real world satellite remote sensing dataset (FMoW) variant of LP-FT gets state-of-the-art accuracies on an NLP continual learning benchmark.
  arXiv  Detail & Related papers  (2023-02-26T10:39:38Z)

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.