# HelpSteer2-Preference: Complementing Ratings with Preferences

arXiv: https://arxiv.org/abs/2410.01257 · HTML: https://mlsanity.com/paper/2410.01257/

## Key facts

- Cited by 130 papers in the map (data through 14 Oct 2026).
- Placed in Reward modeling and RLHF, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 15 of its 34 references that are in the map.
- First posted on arXiv 2 Oct 2024.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Reward modeling and RLHF](https://mlsanity.com/alignment/reward-modeling-and-rlhf/)

## Builds on (most cited first, 15 of 15)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [WebGPT: Browser-assisted question-answering with human feedback](https://mlsanity.com/paper/2112.09332/) (2021)
- [SimPO: Simple Preference Optimization with a Reference-Free Reward](https://mlsanity.com/paper/2405.14734/) (2024)
- [Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluators](https://mlsanity.com/paper/2404.04475/) (2024)
- [Interpretable Preferences via Multi-Objective Reward Modeling and Mixture-of-Experts](https://mlsanity.com/paper/2406.12845/) (2024)
- [Evaluating Large Language Models at Evaluating Instruction Following](https://mlsanity.com/paper/2310.07641/) (2023)
- [Understanding Dataset Difficulty with V-Usable Information](https://mlsanity.com/paper/2110.08420/) (2021)
- [HelpSteer2: Open-source dataset for training top-performing reward models](https://mlsanity.com/paper/2406.08673/) (2024)
- [Arithmetic Control of LLMs for Diverse User Preferences: Directional Preference Alignment with Multi-Objective Rewards](https://mlsanity.com/paper/2402.18571/) (2024)
- [HelpSteer: Multi-attribute Helpfulness Dataset for SteerLM](https://mlsanity.com/paper/2311.09528/) (2023)
- [OffsetBias: Leveraging Debiased Data for Tuning Evaluators](https://mlsanity.com/paper/2407.06551/) (2024)
- [DisCor: Corrective Feedback in Reinforcement Learning via Distribution Correction](https://mlsanity.com/paper/2003.07305/) (2020)
- [Direct Judgement Preference Optimization](https://mlsanity.com/paper/2409.14664/) (2024)
- [Towards Data-Centric RLHF: Simple Metrics for Preference Dataset Comparison](https://mlsanity.com/paper/2409.09603/) (2024)

## Built on by (most cited first, 30 of 130)

- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [Tulu 3: Pushing Frontiers in Open Language Model Post-Training](https://mlsanity.com/paper/2411.15124/) (2024)
- [LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods](https://mlsanity.com/paper/2412.05579/) (2024)
- [LongBench v2: Towards Deeper Understanding and Reasoning on Realistic Long-context Multitasks](https://mlsanity.com/paper/2412.15204/) (2024)
- [Improving Video Generation with Human Feedback](https://mlsanity.com/paper/2501.13918/) (2025)
- [Skywork-Reward-V2: Scaling Preference Data Curation via Human-AI Synergy](https://mlsanity.com/paper/2507.01352/) (2025)
- [RewardBench 2: Advancing Reward Model Evaluation](https://mlsanity.com/paper/2506.01937/) (2025)
- [Checklists Are Better Than Reward Models For Aligning Language Models](https://mlsanity.com/paper/2507.18624/) (2025)
- [Multi-Agent Collaboration via Evolving Orchestration](https://mlsanity.com/paper/2505.19591/) (2025)
- [TRAIL: Trace Reasoning and Agentic Issue Localization](https://mlsanity.com/paper/2505.08638/) (2025)
- [VL-RewardBench: A Challenging Benchmark for Vision-Language Generative Reward Models](https://mlsanity.com/paper/2411.17451/) (2024)
- [Learning to Plan & Reason for Evaluation with Thinking-LLM-as-a-Judge](https://mlsanity.com/paper/2501.18099/) (2025)
- [AceMath: Advancing Frontier Math Reasoning with Post-Training and Reward Modeling](https://mlsanity.com/paper/2412.15084/) (2024)
- [HelpSteer3-Preference: Open Human-Annotated Preference Data across Diverse Tasks and Languages](https://mlsanity.com/paper/2505.11475/) (2025)
- [InternLM-XComposer2.5-Reward: A Simple Yet Effective Multi-Modal Reward Model](https://mlsanity.com/paper/2501.12368/) (2025)
- [Self-Generated Critiques Boost Reward Modeling for Language Models](https://mlsanity.com/paper/2411.16646/) (2024)
- [NoveltyBench: Evaluating Language Models for Humanlike Diversity](https://mlsanity.com/paper/2504.05228/) (2025)
- [DailyDilemmas: Revealing Value Preferences of LLMs with Quandaries of Daily Life](https://mlsanity.com/paper/2410.02683/) (2024)
- [No Free Labels: Limitations of LLM-as-a-Judge Without Human Grounding](https://mlsanity.com/paper/2503.05061/) (2025)
- [Chasing Moving Targets with Online Self-Play Reinforcement Learning for Safer Language Models](https://mlsanity.com/paper/2506.07468/) (2025)
- [Improving LLM-as-a-Judge Inference with the Judgment Distribution](https://mlsanity.com/paper/2503.03064/) (2025)
- [Rethinking External Slow-Thinking: From Snowball Errors to Probability of Correct Reasoning](https://mlsanity.com/paper/2501.15602/) (2025)
- [MicroVQA: A Multimodal Reasoning Benchmark for Microscopy-Based Scientific Research](https://mlsanity.com/paper/2503.13399/) (2025)
- [Open Rubric System: Scaling Reinforcement Learning with Pairwise Adaptive Rubric](https://mlsanity.com/paper/2602.14069/) (2026)
- [Hybrid Preferences: Learning to Route Instances for Human vs. AI Feedback](https://mlsanity.com/paper/2410.19133/) (2024)
- [OpenCodeReasoning-II: A Simple Test Time Scaling Approach via Self-Critique](https://mlsanity.com/paper/2507.09075/) (2025)
- [U-MATH: A University-Level Benchmark for Evaluating Mathematical Skills in LLMs](https://mlsanity.com/paper/2412.03205/) (2024)
- [RAG-RewardBench: Benchmarking Reward Models in Retrieval Augmented Generation for Preference Alignment](https://mlsanity.com/paper/2412.13746/) (2024)
- [R3: Robust Rubric-Agnostic Reward Models](https://mlsanity.com/paper/2505.13388/) (2025)
- [LEMMA: Learning from Errors for MatheMatical Advancement in LLMs](https://mlsanity.com/paper/2503.17439/) (2025)

## Abstract (arXiv)

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