From Outcomes to Processes: Guiding PRM Learning from ORM for Inference-Time Alignment
By: Bin Xie , Bingbing Xu , Yige Yuan and more
Potential Business Impact:
Makes AI understand and follow instructions better.
Inference-time alignment methods have gained significant attention for their efficiency and effectiveness in aligning large language models (LLMs) with human preferences. However, existing dominant approaches using reward-guided search (RGS) primarily rely on outcome reward models (ORMs), which suffer from a critical granularity mismatch: ORMs are designed to provide outcome rewards for complete responses, while RGS methods rely on process rewards to guide the policy, leading to inconsistent scoring and suboptimal alignment. To address this challenge, we introduce process reward models (PRMs) into RGS and argue that an ideal PRM should satisfy two objectives: Score Consistency, ensuring coherent evaluation across partial and complete responses, and Preference Consistency, aligning partial sequence assessments with human preferences. Based on these, we propose SP-PRM, a novel dual-consistency framework integrating score consistency-based and preference consistency-based partial evaluation modules without relying on human annotation. Extensive experiments on dialogue, summarization, and reasoning tasks demonstrate that SP-PRM substantially enhances existing RGS methods, achieving a 3.6%-10.3% improvement in GPT-4 evaluation scores across all tasks.
Similar Papers
A Survey of Process Reward Models: From Outcome Signals to Process Supervisions for Large Language Models
Computation and Language
Teaches computers to think step-by-step.
The Bidirectional Process Reward Model
Computation and Language
Helps AI check its thinking both ways.
Beyond Correctness: Harmonizing Process and Outcome Rewards through RL Training
Machine Learning (CS)
Teaches computers to reason better, step-by-step.