PyBangla at BLP-2025 Task 2: Enhancing Bangla-to-Python Code Generation with Iterative Self-Correction and Multilingual Agents

TL;DR

This paper introduces BanglaCodeAct, a novel agent-based framework utilizing multi-agent prompting and iterative self-correction to significantly improve Bangla-to-Python code generation with open-source multilingual LLMs, setting new benchmarks.

Contribution

It presents BanglaCodeAct, an open-source, agent-based approach that enhances low-resource Bangla code generation without task-specific fine-tuning, using a Thought-Code-Observation loop with multilingual LLMs.

Findings

Qwen3-8B with BanglaCodeAct achieves 94.0% pass@1 on dev set.

Achieves 71.6% pass@1 on blind test set.

Establishes new benchmarks for Bangla-to-Python translation.

Abstract

LLMs excel at code generation from English prompts, but this progress has not extended to low-resource languages. We address Bangla-to-Python code generation by introducing BanglaCodeAct, an agent-based framework that leverages multi-agent prompting and iterative self-correction. Unlike prior approaches relying on task-specific fine-tuning, BanglaCodeAct employs an open-source multilingual LLM within a Thought-Code-Observation loop, enabling dynamic generation, testing, and refinement of code from Bangla instructions. We benchmark several small-parameter open-source LLMs and evaluate their effectiveness on the mHumanEval dataset for Bangla NL2Code. Our results show that Qwen3-8B, when deployed with BanglaCodeAct, achieves the best performance, with pass@1 accuracy of 94.0\% on the development set and 71.6\% on the blind test set. These results establish a new benchmark for Bangla-to-Python translation and highlight the potential of agent-based reasoning for reliable code generation in low-resource languages. Experimental scripts are publicly available at github.com/jahidulzaid/PyBanglaCodeActAgent.