Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Gurjar, Priya, Ishmam, Md Farhan, Marino, Kenneth
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2604.17244
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866910145760460800
author Gurjar, Priya
Ishmam, Md Farhan
Marino, Kenneth
author_facet Gurjar, Priya
Ishmam, Md Farhan
Marino, Kenneth
contents Despite the rapid progress, LLMs for sequential decision-making (i.e., LLM agents) still struggle to produce diverse outputs. This leads to insufficient exploration, convergence to sub-optimal solutions, and becoming stuck in loops. Such limitations can be problematic in environments that require active exploration to gather information and make decisions. Sampling methods such as temperature scaling introduce token-level randomness but fail to produce enough diversity at the sequence level. We analyze LLM exploration in the classic Multi-Armed Bandit (MAB) setting and the Text Adventure Learning Environment Suite (TALES). We find that current decoding strategies and prompting methods like Chain-of-Thought and Tree-of-Thought are insufficient for robust exploration. To address this, we introduce DORA Explorer (Diversity-Oriented Ranking of Actions), a training-free framework for improving exploration in LLM agents. DORA generates diverse action candidates, scores them using token log-probabilities, and selects actions using a tunable exploration parameter. DORA achieves UCB-competitive performance on MAB and consistent gains across TALES, e.g., improving Qwen2.5-7B's performance from 29.2% to 45.5% in TextWorld. Our project is available at: https://dora-explore.github.io/.
format Preprint
id arxiv_https___arxiv_org_abs_2604_17244
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle DORA Explorer: Improving the Exploration Ability of LLMs Without Training
Gurjar, Priya
Ishmam, Md Farhan
Marino, Kenneth
Computation and Language
Artificial Intelligence
Despite the rapid progress, LLMs for sequential decision-making (i.e., LLM agents) still struggle to produce diverse outputs. This leads to insufficient exploration, convergence to sub-optimal solutions, and becoming stuck in loops. Such limitations can be problematic in environments that require active exploration to gather information and make decisions. Sampling methods such as temperature scaling introduce token-level randomness but fail to produce enough diversity at the sequence level. We analyze LLM exploration in the classic Multi-Armed Bandit (MAB) setting and the Text Adventure Learning Environment Suite (TALES). We find that current decoding strategies and prompting methods like Chain-of-Thought and Tree-of-Thought are insufficient for robust exploration. To address this, we introduce DORA Explorer (Diversity-Oriented Ranking of Actions), a training-free framework for improving exploration in LLM agents. DORA generates diverse action candidates, scores them using token log-probabilities, and selects actions using a tunable exploration parameter. DORA achieves UCB-competitive performance on MAB and consistent gains across TALES, e.g., improving Qwen2.5-7B's performance from 29.2% to 45.5% in TextWorld. Our project is available at: https://dora-explore.github.io/.
title DORA Explorer: Improving the Exploration Ability of LLMs Without Training
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2604.17244