HelpSteer2-Preference: Complementing Ratings with Preferences

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Zhilin, Bukharin, Alexander, Delalleau, Olivier, Egert, Daniel, Shen, Gerald, Zeng, Jiaqi, Kuchaiev, Oleksii, Dong, Yi
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910860651266048
author Wang, Zhilin
Bukharin, Alexander
Delalleau, Olivier
Egert, Daniel
Shen, Gerald
Zeng, Jiaqi
Kuchaiev, Oleksii
Dong, Yi
author_facet Wang, Zhilin
Bukharin, Alexander
Delalleau, Olivier
Egert, Daniel
Shen, Gerald
Zeng, Jiaqi
Kuchaiev, Oleksii
Dong, Yi
contents Reward models are critical for aligning models to follow instructions, and are typically trained following one of two popular paradigms: Bradley-Terry style or Regression style. However, there is a lack of evidence that either approach is better than the other, when adequately matched for data. This is primarily because these approaches require data collected in different (but incompatible) formats, meaning that adequately matched data is not available in existing public datasets. To tackle this problem, we release preference annotations (designed for Bradley-Terry training) to complement existing ratings (designed for Regression style training) in the HelpSteer2 dataset. To improve data interpretability, preference annotations are accompanied with human-written justifications. Using this data, we conduct the first head-to-head comparison of Bradley-Terry and Regression models when adequately matched for data. Based on insights derived from such a comparison, we propose a novel approach to combine Bradley-Terry and Regression reward modeling. A Llama-3.1-70B-Instruct model tuned with this approach scores 94.1 on RewardBench, emerging top of more than 140 reward models as of 1 Oct 2024. This reward model can then be used with REINFORCE algorithm (RLHF) to align an Instruct model to reach 85.0 on Arena Hard, which is No. 1 as of 1 Oct 2024. We open-source this dataset (CC-BY-4.0 license) at https://huggingface.co/datasets/nvidia/HelpSteer2#preferences-new -- 1-oct-2024 and openly release the trained Reward and Instruct models at https://huggingface.co/nvidia/Llama-3.1-Nemotron-70B-Reward and https://huggingface.co/nvidia/Llama-3.1-Nemotron-70B-Instruct
format Preprint
id arxiv_https___arxiv_org_abs_2410_01257
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle HelpSteer2-Preference: Complementing Ratings with Preferences
Wang, Zhilin
Bukharin, Alexander
Delalleau, Olivier
Egert, Daniel
Shen, Gerald
Zeng, Jiaqi
Kuchaiev, Oleksii
Dong, Yi
Machine Learning
Artificial Intelligence
Computation and Language
Reward models are critical for aligning models to follow instructions, and are typically trained following one of two popular paradigms: Bradley-Terry style or Regression style. However, there is a lack of evidence that either approach is better than the other, when adequately matched for data. This is primarily because these approaches require data collected in different (but incompatible) formats, meaning that adequately matched data is not available in existing public datasets. To tackle this problem, we release preference annotations (designed for Bradley-Terry training) to complement existing ratings (designed for Regression style training) in the HelpSteer2 dataset. To improve data interpretability, preference annotations are accompanied with human-written justifications. Using this data, we conduct the first head-to-head comparison of Bradley-Terry and Regression models when adequately matched for data. Based on insights derived from such a comparison, we propose a novel approach to combine Bradley-Terry and Regression reward modeling. A Llama-3.1-70B-Instruct model tuned with this approach scores 94.1 on RewardBench, emerging top of more than 140 reward models as of 1 Oct 2024. This reward model can then be used with REINFORCE algorithm (RLHF) to align an Instruct model to reach 85.0 on Arena Hard, which is No. 1 as of 1 Oct 2024. We open-source this dataset (CC-BY-4.0 license) at https://huggingface.co/datasets/nvidia/HelpSteer2#preferences-new -- 1-oct-2024 and openly release the trained Reward and Instruct models at https://huggingface.co/nvidia/Llama-3.1-Nemotron-70B-Reward and https://huggingface.co/nvidia/Llama-3.1-Nemotron-70B-Instruct
title HelpSteer2-Preference: Complementing Ratings with Preferences
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2410.01257