Evaluating LLMs' Multilingual Capabilities for Bengali: Benchmark Creation and Performance Analysis

TL;DR

This paper creates a benchmark to evaluate Bengali language capabilities in large language models, revealing performance gaps, error patterns, and the impact of tokenization, thus guiding future improvements for underrepresented languages.

Contribution

It introduces a standardized benchmark for Bengali NLP, evaluates 10 LLMs, and provides insights into their strengths and weaknesses specific to Bengali language processing.

Findings

Smaller models and certain architectures like Mistral underperform in Bengali.

Robustness varies across models, with DeepSeek showing stable multilingual performance.

Tokenization efficiency inversely affects model accuracy, impacting Bengali NLP performance.

Abstract

Bengali is an underrepresented language in NLP research. However, it remains a challenge due to its unique linguistic structure and computational constraints. In this work, we systematically investigate the challenges that hinder Bengali NLP performance by focusing on the absence of standardized evaluation benchmarks. We then evaluated 10 recent open source Large Language Models (LLMs) in 8 of the translated datasets and performed a comprehensive error analysis to pinpoint their primary failure modes. Our findings reveal consistent performance gaps for Bengali compared to English, particularly for smaller models and specific model families like Mistral. We also identified promising robustness in certain architectures, such as DeepSeek, that maintain more stable performance across languages. Our analysis reveals an inverse relationship between tokenization efficiency and LLM accuracy where models tend to perform worse when inputs are excessively tokenized, whereas more efficient \& concise tokenization results in improved performance. These findings highlight critical areas where current models fall short and underscore the need for improved dataset quality and evaluation methodologies tailored to multilingual contexts. This work will catalyze further research on NLP for underrepresented languages, helping to democratize access to advanced language technologies worldwide. The code and dataset used in this research is publicly available at https://github.com/BengaliAI/bn-llm-benchmark.