WildReward: Learning Reward Models from In-the-Wild Human Interactions
This addresses the need for scalable reward modeling in AI by leveraging abundant real-world data, though it is incremental as it builds on existing reward model frameworks.
The paper tackles the problem of training reward models for large language models by using in-the-wild human interactions instead of annotated preference pairs, resulting in WildReward achieving comparable or superior performance with improved calibration and consistency, and showing significant improvements when applied to online DPO training.
Reward models (RMs) are crucial for the training of large language models (LLMs), yet they typically rely on large-scale human-annotated preference pairs. With the widespread deployment of LLMs, in-the-wild interactions have emerged as a rich source of implicit reward signals. This raises the question: Can we develop reward models directly from in-the-wild interactions? In this work, we explore this possibility by adopting WildChat as an interaction source and proposing a pipeline to extract reliable human feedback, yielding 186k high-quality instances for training WildReward via ordinal regression directly on user feedback without preference pairs. Extensive experiments demonstrate that WildReward achieves comparable or even superior performance compared to conventional reward models, with improved calibration and cross-sample consistency. We also observe that WildReward benefits directly from user diversity, where more users yield stronger reward models. Finally, we apply WildReward to online DPO training and observe significant improvements across various tasks. Code and data are released at https://github.com/THU-KEG/WildReward.