Positive-Unlabeled Reinforcement Learning Distillation for On-Premise Small Models

TL;DR

This paper introduces a positive-unlabeled reinforcement learning distillation approach that enables on-premise small models to learn preferences without human labels or large reward models, reducing costs and maintaining performance.

Contribution

It presents a novel PU RL distillation method that distills preference-optimization capabilities into small models using only black-box teacher responses, suitable for on-premise deployment.

Findings

Achieves strong performance with low-cost queries.

Effectively distills preference signals without human labels.

Theoretically justified stability and near-optimal focus.

Abstract

Due to constraints on privacy, cost, and latency, on-premise deployment of small models is increasingly common. However, most practical pipelines stop at supervised fine-tuning (SFT) and fail to reach the reinforcement learning (RL) alignment stage. The main reason is that RL alignment typically requires either expensive human preference annotation or heavy reliance on high-quality reward models with large-scale API usage and ongoing engineering maintenance, both of which are ill-suited to on-premise settings. To bridge this gap, we propose a positive-unlabeled (PU) RL distillation method for on-premise small-model deployment. Without human-labeled preferences or a reward model, our method distills the teacher's preference-optimization capability from black-box generations into a locally trainable student. For each prompt, we query the teacher once to obtain an anchor response, locally sample multiple student candidates, and perform anchor-conditioned self-ranking to induce pairwise or listwise preferences, enabling a fully local training loop via direct preference optimization or group relative policy optimization. Theoretical analysis justifies that the induced preference signal by our method is order-consistent and concentrates on near-optimal candidates, supporting its stability for preference optimization. Experiments demonstrate that our method achieves consistently strong performance under a low-cost setting.