Saved in:
Bibliographic Details
Main Authors: Rietz, Finn, Smirnov, Oleg, Karimi, Sara, Cao, Lele
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2502.04979
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915526577487872
author Rietz, Finn
Smirnov, Oleg
Karimi, Sara
Cao, Lele
author_facet Rietz, Finn
Smirnov, Oleg
Karimi, Sara
Cao, Lele
contents Prompt tuning has emerged as a key technique for adapting large pre-trained Decision Transformers (DTs) in offline Reinforcement Learning (RL), particularly in multi-task and few-shot settings. The Prompting Decision Transformer (PDT) enables task generalization via trajectory prompts sampled uniformly from expert demonstrations -- without accounting for prompt informativeness. In this work, we propose a bandit-based prompt-tuning method that learns to construct optimal trajectory prompts from demonstration data at inference time. We devise a structured bandit architecture operating in the trajectory prompt space, achieving linear rather than combinatorial scaling with prompt size. Additionally, we show that the pre-trained PDT itself can serve as a powerful feature extractor for the bandit, enabling efficient reward modeling across various environments. We theoretically establish regret bounds and demonstrate empirically that our method consistently enhances performance across a wide range of tasks, high-dimensional environments, and out-of-distribution scenarios, outperforming existing baselines in prompt tuning.
format Preprint
id arxiv_https___arxiv_org_abs_2502_04979
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Prompt Tuning Decision Transformers with Structured and Scalable Bandits
Rietz, Finn
Smirnov, Oleg
Karimi, Sara
Cao, Lele
Machine Learning
Prompt tuning has emerged as a key technique for adapting large pre-trained Decision Transformers (DTs) in offline Reinforcement Learning (RL), particularly in multi-task and few-shot settings. The Prompting Decision Transformer (PDT) enables task generalization via trajectory prompts sampled uniformly from expert demonstrations -- without accounting for prompt informativeness. In this work, we propose a bandit-based prompt-tuning method that learns to construct optimal trajectory prompts from demonstration data at inference time. We devise a structured bandit architecture operating in the trajectory prompt space, achieving linear rather than combinatorial scaling with prompt size. Additionally, we show that the pre-trained PDT itself can serve as a powerful feature extractor for the bandit, enabling efficient reward modeling across various environments. We theoretically establish regret bounds and demonstrate empirically that our method consistently enhances performance across a wide range of tasks, high-dimensional environments, and out-of-distribution scenarios, outperforming existing baselines in prompt tuning.
title Prompt Tuning Decision Transformers with Structured and Scalable Bandits
topic Machine Learning
url https://arxiv.org/abs/2502.04979