Dual-Weighted Reinforcement Learning for Generative Preference Modeling
This addresses a gap in applying RL to general preference modeling tasks beyond verifiable domains, offering a framework for reasoning-enhanced preference learning.
The paper tackles the challenge of extending reinforcement learning to non-verifiable tasks using human preference pairs by proposing Dual-Weighted Reinforcement Learning (DWRL), which integrates chain-of-thought reasoning with the Bradley-Terry model and outperforms baselines across multiple benchmarks and model scales like Llama3 and Qwen2.5.
Reinforcement learning (RL) has recently proven effective at scaling chain-of-thought (CoT) reasoning in large language models on tasks with verifiable answers. However, extending RL to more general non-verifiable tasks, typically in the format of human preference pairs, remains both challenging and underexplored. In this work, we propose Dual-Weighted Reinforcement Learning (DWRL), a new framework for preference modeling that integrates CoT reasoning with the Bradley-Terry (BT) model via a dual-weighted RL objective that preserves preference-modeling inductive bias. DWRL approximates the maximum-likelihood objective of the BT model with two complementary weights: an instance-wise misalignment weight, which emphasizes under-trained pairs misaligned with human preference, and a group-wise (self-normalized) conditional preference score, which promotes promising thoughts. In this paper, we apply DWRL to preference modeling by training generative preference models (GPMs) to first generate a thought and then predict the human preference score. Across multiple benchmarks and model scales (Llama3 and Qwen2.5), DWRL consistently outperforms both GPM baselines and scalar models, while producing coherent, interpretable thoughts. In summary, our results position DWRL as a general framework for reasoning-enhanced preference learning beyond verifiable tasks.