High-quality preference datasets are essential for training reward models that can effectively guide large language models (LLMs) in generating high-quality responses aligned with human preferences.As LLMs become stronger and better aligned, permissively licensed preference datasets, such as Open Assistant, HH-RLHF, and HelpSteer need to be updated to remain effective for reward modeling.Methods that distil preference data from proprietary LLMs such as GPT-4 have restrictions on commercial usage imposed by model providers.To improve upon both generated responses and attribute labeling quality, we release HelpSteer2, a permissively licensed preference dataset (CC-BY-4.0). Using a powerful Nemotron-4-340B base model trained on HelpSteer2, we are able to achieve the SOTA score (92.0%) on Reward-Bench's primary dataset, outperforming currently listed open and proprietary models, as of June 12th, 2024.Notably, HelpSteer2 consists of only ten thousand response pairs, an order of magnitude fewer than existing preference datasets (e.g., HH-RLHF), which makes it highly efficient for training reward models. Our extensive experiments demonstrate that reward models trained with HelpSteer2 are effective in aligning LLMs. Additionally, we propose SteerLM 2.0, a model alignment approach that can effectively make use of the rich multi-attribute score predicted by our reward models.

Large language models (LLMs) are currently aligned using techniques such as reinforcement learning from human feedback (RLHF). However, these methods use scalar rewards that can only reflect user preferences on average . Pluralistic alignment instead seeks to capture diverse user preferences across a set of attributes, moving beyond just helpfulness and harmlessness. Toward this end, we propose a steerable pluralistic model based on few-shot comparative regression that can adapt to individual user preferences. Our approach leverages in-context learning and reasoning, grounded in a set of fine-grained attributes, to compare response options and make aligned choices. To evaluate our algorithm, we also propose two new steerable pluralistic benchmarks by adapting the Moral Integrity Corpus (MIC) and the HelpSteer2 datasets, demonstrating the applicability of our approach to value-aligned decision-making and reward modeling, respectively. Our few-shot comparative regression approach is interpretable and compatible with different attributes and LLMs, while outperforming multiple baseline and state-of-the-art methods. Our work provides new insights and research directions in pluralistic alignment, enabling a more fair and representative use of LLMs and advancing the state-of-the-art in ethical AI.

High-quality preference datasets are essential for training reward models that can effectively guide large language models (LLMs) in generating high-quality responses aligned with human preferences.As LLMs become stronger and better aligned, permissively licensed preference datasets, such as Open Assistant, HH-RLHF, and HelpSteer need to be updated to remain effective for reward modeling.Methods that distil preference data from proprietary LLMs such as GPT-4 have restrictions on commercial usage imposed by model providers.To improve upon both generated responses and attribute labeling quality, we release HelpSteer2, a permissively licensed preference dataset (CC-BY-4.0). Using a powerful Nemotron-4-340B base model trained on HelpSteer2, we are able to achieve the SOTA score (92.0%) on Reward-Bench's primary dataset, outperforming currently listed open and proprietary models, as of June 12th, 2024.Notably, HelpSteer2 consists of only ten thousand response pairs, an order of magnitude fewer than existing preference datasets (e.g., HH-RLHF), which makes it highly efficient for training reward models. Our extensive experiments demonstrate that reward models trained with HelpSteer2 are effective in aligning LLMs. Additionally, we propose SteerLM 2.0, a model alignment approach that can effectively make use of the rich multi-attribute score predicted by our reward models.

Learning from human feedback has enabled the alignment of language models (LMs) with human preferences. However, directly collecting human preferences can be expensive, time-consuming, and can have high variance. An appealing alternative is to distill preferences from LMs as a source of synthetic annotations as they are more consistent, cheaper, and scale better than human annotation; however, they are also prone to biases and errors. In this work, we introduce a routing framework that combines inputs from humans and LMs to achieve better annotation quality, while reducing the total cost of human annotation. The crux of our approach is to identify preference instances that will benefit from human annotations. We formulate this as an optimization problem: given a preference dataset and an evaluation metric, we train a performance prediction model to predict a reward model's performance on an arbitrary combination of human and LM annotations and employ a routing strategy that selects a combination that maximizes predicted performance. We train the performance prediction model on MultiPref, a new preference dataset with 10K instances paired with human and LM labels. We show that the selected hybrid mixture of LM and direct human preferences using our routing framework achieves better reward model performance compared to using either one exclusively. We simulate selective human preference collection on three other datasets and show that our method generalizes well to all three. We analyze features from the routing model to identify characteristics of instances that can benefit from human feedback, e.g., prompts with a moderate safety concern or moderate intent complexity. We release the dataset, annotation platform, and source code used in this study to foster more efficient and accurate preference collection in the future.

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. We also demonstrate the effectiveness of this reward model at aligning models to follow instructions in RLHF. We open-source this dataset (CC-BY-4.0 license) at https://huggingface.co/datasets/nvidia/HelpSteer2 and openly release the trained Reward Model at https://huggingface.co/nvidia/Llama-3.1-Nemotron-70B-Reward

High-quality preference datasets are essential for training reward models that can effectively guide large language models (LLMs) in generating high-quality responses aligned with human preferences. As LLMs become stronger and better aligned, permissively licensed preference datasets, such as Open Assistant, HH-RLHF, and HelpSteer need to be updated to remain effective for reward modeling. Methods that distil preference data from proprietary LLMs such as GPT-4 have restrictions on commercial usage imposed by model providers. To improve upon both generated responses and attribute labeling quality, we release HelpSteer2, a permissively licensed preference dataset (CC-BY-4.0). Using a powerful internal base model trained on HelpSteer2, we are able to achieve the SOTA score (92.0%) on Reward-Bench's primary dataset, outperforming currently listed open and proprietary models, as of June 12th, 2024. Notably, HelpSteer2 consists of only ten thousand response pairs, an order of magnitude fewer than existing preference datasets (e.g., HH-RLHF), which makes it highly efficient for training reward models. Our extensive experiments demonstrate that reward models trained with HelpSteer2 are effective in aligning LLMs. In particular, we propose SteerLM 2.0, a model alignment approach that can effectively make use of the rich multi-attribute score predicted by our reward models. HelpSteer2 is available at https://huggingface.co/datasets/nvidia/HelpSteer2 and code is available at https://github.com/NVIDIA/NeMo-Aligner