English

PREFINE: Preference-Based Implicit Reward and Cost Fine-Tuning for Safety Alignment

Machine Learning 2026-05-21 v1 Artificial Intelligence

Abstract

We address the problem of making a pre-trained reinforcement learning (RL) policy safety-aware by incorporating cost constraints without retraining it from scratch. While costs could be numerically encoded, we assume a more general setting is when costs are provided as preferences. Given a reward-optimized policy and a small dataset of preferred (low-cost) and dispreferred (high-cost) trajectories, our goal is to fine-tune the policy to generate low-cost behaviors while retaining high rewards. Unlike standard RLHF in language models, where preferences are defined over responses to the same prompt, our setting involves trajectory-level preferences in continuous control environments. We introduce PREFINE: Preference-based Implicit Reward and Cost Fine-Tuning for Safety Alignment which is a preference-based fine-tuning method that adapts Direct Preference Optimization (DPO), which is now widely used for LLM fine-tuning, to the sequential decision making setting. PREFINE constructs policy-sampled counterfactual trajectories to establish meaningful preference contrasts and jointly optimizes for reward retention and safety alignment. Empirically, PREFINE reduces constraint violations and catastrophic failures by over 60% while maintaining original reward behavior. PREFINE produces policies that achieve low-cost, high-reward performance with significantly improved data and computational efficiency compared to full offline RL or imitation learning, bridging preference alignment and safe policy adaptation in continuous domains.

Keywords

Cite

@article{arxiv.2605.21225,
  title  = {PREFINE: Preference-Based Implicit Reward and Cost Fine-Tuning for Safety Alignment},
  author = {Richa Verma and Bavish Kulur and Sanjay Chawla and Balaraman Ravindran},
  journal= {arXiv preprint arXiv:2605.21225},
  year   = {2026}
}

Comments

Accepted at AAMAS 2026 as a full paper

R2 v1 2026-07-22T07:24:07.196Z