AILGDec 12, 2024

In-Dataset Trajectory Return Regularization for Offline Preference-based Reinforcement Learning

arXiv:2412.09104v25 citationsh-index: 26AAAI
Originality Incremental advance
AI Analysis

This work addresses a specific challenge in offline PbRL for improving policy learning from human preferences, representing an incremental advancement.

The paper tackles the problem of reward bias and optimistic trajectory stitching in offline preference-based reinforcement learning by proposing In-Dataset Trajectory Return Regularization (DTR), which uses conditional sequence modeling to balance behavior policy fidelity and reward optimization, achieving superior performance on benchmarks.

Offline preference-based reinforcement learning (PbRL) typically operates in two phases: first, use human preferences to learn a reward model and annotate rewards for a reward-free offline dataset; second, learn a policy by optimizing the learned reward via offline RL. However, accurately modeling step-wise rewards from trajectory-level preference feedback presents inherent challenges. The reward bias introduced, particularly the overestimation of predicted rewards, leads to optimistic trajectory stitching, which undermines the pessimism mechanism critical to the offline RL phase. To address this challenge, we propose In-Dataset Trajectory Return Regularization (DTR) for offline PbRL, which leverages conditional sequence modeling to mitigate the risk of learning inaccurate trajectory stitching under reward bias. Specifically, DTR employs Decision Transformer and TD-Learning to strike a balance between maintaining fidelity to the behavior policy with high in-dataset trajectory returns and selecting optimal actions based on high reward labels. Additionally, we introduce an ensemble normalization technique that effectively integrates multiple reward models, balancing the tradeoff between reward differentiation and accuracy. Empirical evaluations on various benchmarks demonstrate the superiority of DTR over other state-of-the-art baselines.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes