TGPR: Tree-Guided Policy Refinement for Robust Self-Debugging of LLMs

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Ozerova, Daria, Trofimova, Ekaterina
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866909831122649088
author Ozerova, Daria
Trofimova, Ekaterina
author_facet Ozerova, Daria
Trofimova, Ekaterina
contents Iterative refinement has been a promising paradigm to enable large language models (LLMs) to resolve difficult reasoning and problem-solving tasks. One of the key challenges, however, is how to effectively search through the enormous search space of possible refinements. Existing methods typically fall back on predefined heuristics, which are troubled by the exploration-exploitation dilemma and cannot adapt based on past refinement outcomes. We introduce Tree-Guided Policy Refinement (TGPR), a novel framework that combines GRPO with a Thompson-Sampling-based tree search. TGPR explores both failed and successful refinement paths actively, with denser training trajectories and more adaptive policies. On HumanEval, MBPP, and APPS benchmarks, our method achieves up to +4.2 percentage points absolute improvement in pass@1 (on MBPP) and up to +12.51 percentage points absolute improvement in pass@10 (on APPS) compared to a competitive GRPO baseline. Apart from debugging code, TGPR focuses on a principled approach to combining learned policies with structured search methods, offering a general framework for enhancing iterative refinement and stateful reasoning in LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2510_06878
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle TGPR: Tree-Guided Policy Refinement for Robust Self-Debugging of LLMs
Ozerova, Daria
Trofimova, Ekaterina
Artificial Intelligence
Iterative refinement has been a promising paradigm to enable large language models (LLMs) to resolve difficult reasoning and problem-solving tasks. One of the key challenges, however, is how to effectively search through the enormous search space of possible refinements. Existing methods typically fall back on predefined heuristics, which are troubled by the exploration-exploitation dilemma and cannot adapt based on past refinement outcomes. We introduce Tree-Guided Policy Refinement (TGPR), a novel framework that combines GRPO with a Thompson-Sampling-based tree search. TGPR explores both failed and successful refinement paths actively, with denser training trajectories and more adaptive policies. On HumanEval, MBPP, and APPS benchmarks, our method achieves up to +4.2 percentage points absolute improvement in pass@1 (on MBPP) and up to +12.51 percentage points absolute improvement in pass@10 (on APPS) compared to a competitive GRPO baseline. Apart from debugging code, TGPR focuses on a principled approach to combining learned policies with structured search methods, offering a general framework for enhancing iterative refinement and stateful reasoning in LLMs.
title TGPR: Tree-Guided Policy Refinement for Robust Self-Debugging of LLMs
topic Artificial Intelligence
url https://arxiv.org/abs/2510.06878