Learning in Context, Guided by Choice: A Reward-Free Paradigm for Reinforcement Learning with Transformers

TL;DR

This paper introduces a new paradigm for reinforcement learning that relies solely on preference feedback instead of explicit rewards, enabling effective in-context learning and generalization without reward supervision.

Contribution

It proposes ICPRL, a preference-based reinforcement learning framework that eliminates the need for reward signals, and demonstrates its effectiveness across various tasks.

Findings

ICPRL achieves comparable performance to reward-based methods.

Supervised pretraining remains effective with preference-only data.

Preference-native frameworks improve data efficiency.

Abstract

In-context reinforcement learning (ICRL) leverages the in-context learning capabilities of transformer models (TMs) to efficiently generalize to unseen sequential decision-making tasks without parameter updates. However, existing ICRL methods rely on explicit reward signals during pretraining, which limits their applicability when rewards are ambiguous, hard to specify, or costly to obtain. To overcome this limitation, we propose a new learning paradigm, In-Context Preference-based Reinforcement Learning (ICPRL), in which both pretraining and deployment rely solely on preference feedback, eliminating the need for reward supervision. We study two variants that differ in the granularity of feedback: Immediate Preference-based RL (I-PRL) with per-step preferences, and Trajectory Preference-based RL (T-PRL) with trajectory-level comparisons. We first show that supervised pretraining, a standard approach in ICRL, remains effective under preference-only context datasets, demonstrating the feasibility of in-context reinforcement learning using only preference signals. To further improve data efficiency, we introduce alternative preference-native frameworks for I-PRL and T-PRL that directly optimize TM policies from preference data without requiring reward signals nor optimal action labels.Experiments on dueling bandits, navigation, and continuous control tasks demonstrate that ICPRL enables strong in-context generalization to unseen tasks, achieving performance comparable to ICRL methods trained with full reward supervision.