Daria Ozerova | Searcharxiv

arXiv2025

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

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.

cs.AI↗

arXiv2025

ML2B: Benchmarking LLMs on Cross-Lingual ML Pipeline Generation

We introduce ML2B, the first benchmark for evaluating cross-lingual task comprehension in end-to-end ML pipeline generation by large language models. Despite growing global AI adoption, no systematic evaluation exists for ML pipeline generation beyond English task descriptions. ML2B addresses this gap with 35 Kaggle competitions spanning tabular, text, and image domains, translated into 14 languages by native-speaker researchers with ML expertise, yielding 490 task-language pairs. To ensure evaluation integrity, the benchmark incorporates 10 private competitions without publicly available solutions and employs network-isolated evaluation infrastructure restricting runtime access to essential ML resources. We provide standardized evaluation protocols, an AutoGluon algorithmic baseline, and comprehensive failure mode analysis. Experiments with frontier models (GPT-4.1-mini, GPT-OSS-120b, Gemini-2.5-Flash) reveal that cross-lingual performance degradation is highly task-dependent rather than following traditional resource-availability hierarchies, with gaps ranging from language advantages to severe degradation depending on competition characteristics. These findings challenge conventional assumptions about multilingual model capabilities and underscore the necessity of systematic cross-lingual evaluation for ML pipeline generation. We open-source the benchmark, baselines, and evaluation infrastructure at https://github.com/enaix/ml2b.

cs.CL↗