BenHalluEval: A Multi-Task Hallucination Evaluation Framework for Large Language Models on Bengali

2026-05-29

AI Summary

This study addresses the lack of systematic evaluation of hallucinations in large language models (LLMs) for Bengali. It proposes the first fine-grained, multitask hallucination benchmark for Bengali, BenHalluBench, comprising 12,000 samples spanning four task categories—generative question answering, Bengali-English code-mixed QA, summarization, and reasoning—and covering 12 task-specific hallucination types. The work introduces a novel dual-track evaluation protocol and a calibrated metric, BenHalluScore, which jointly accounts for false positive rate and hallucination detection rate, while integrating three existing datasets. Experiments across seven mainstream models reveal BenHalluScore ranging from 7.72% to 55.42%, highlighting substantial disparities in hallucination calibration capabilities. Furthermore, chain-of-thought prompting fails to consistently improve discrimination performance, underscoring the limitations of single-track evaluation and prompt-only mitigation strategies.

Abstract

Despite Bengali being the sixth most spoken language in the world, no prior work has systematically evaluated hallucination in large language models (LLMs) for Bengali. We introduce BenHalluEval, a fine-grained hallucination evaluation framework for Bengali covering four tasks: Generative Question Answering (GQA), Bangla-English Code-Mixed QA, Summarization, and Reasoning. We construct 12,000 hallucinated candidates using GPT-5.4 across twelve task-specific hallucination types, drawn from three existing Bengali datasets, and evaluate seven LLMs spanning reasoning-oriented, multilingual, and Bengali-centric categories under a dual-track protocol that independently measures false-positive rate on ground-truth instances (Track A) and hallucination detection rate on hallucinated candidates (Track B). To jointly penalise both failure modes and prevent inflated scores from uniform response bias, we propose BenHalluScore, a dual-track calibration metric that ranges from 7.72% to 55.42% across models and tasks, revealing substantial variation in hallucination calibration. Chain-of-thought prompting, applied as a mitigation strategy, shifts response distributions without consistently improving hallucination discrimination. BenHalluEval establishes the first dedicated hallucination benchmark for Bengali and highlights the inadequacy of single-track and prompting-only evaluation approaches for low-resource language settings. The dataset and code are available at https://anonymous.4open.science/r/BanglaHalluEval-EB77.

Problem

• hallucination
• Bengali
• large language models
• evaluation framework
• low-resource languages

Innovation

• hallucination evaluation
• Bengali LLMs
• multi-task benchmark
• dual-track calibration
• low-resource languages