PRIMT: Preference-based Reinforcement Learning with Multimodal Feedback and Trajectory Synthesis from Foundation Models

Ruiqi Wang, Dezhong Zhao, Ziqin Yuan, Tianyu Shao, Guohua Chen, Dominic Kao, Sungeun Hong, Byung-Cheol Min

Purdue University· Beijing University of Chemical Technology· University of Illinois Urbana-Champaign· Sungkyunkwan University· Indiana University Bloomington

Preference-based Reinforcement Learning Foundation Models for Robotics Neuro-Symbolic Fusion Multimodal Feedback Causal Inference Trajectory Synthesis Robot Manipulation

⋅ NeurIPS ⋅ OpenReview ⋅Oral ⋅Oral 2C Reinforcement/State-space 1

Abstract

Preference-based reinforcement learning (PbRL) has emerged as a promising paradigm for teaching robots complex behaviors without reward engineering. However, its effectiveness is often limited by two critical challenges: the reliance on extensive human input and the inherent difficulties in resolving query ambiguity and credit assignment during reward learning.

In this paper, we introduce PRIMT, a PbRL framework designed to overcome these challenges by leveraging foundation models (FMs) for multimodal synthetic feedback and trajectory synthesis. Unlike prior approaches that rely on single-modality FM evaluations, PRIMT employs a hierarchical neuro-symbolic fusion strategy, integrating the complementary strengths of vision-language models (VLMs) and large language models (LLMs) in evaluating robot behaviors for more reliable and comprehensive feedback.

PRIMT also incorporates foresight trajectory generation to warm-start the trajectory buffer with bootstrapped samples, reducing early-stage query ambiguity, and hindsight trajectory augmentation for counterfactual reasoning with a causal auxiliary loss to improve credit assignment.

We evaluate PRIMT on 2 locomotion and 6 manipulation tasks on various benchmarks, demonstrating superior performance over FM-based and scripted baselines. Website at https://primt25.github.io/.