BengaliFig: A Low-Resource Challenge for Figurative and Culturally Grounded Reasoning in Bengali

TL;DR

BengaliFig introduces a challenging dataset of Bengali riddles to evaluate large language models' abilities in figurative and culturally grounded reasoning within low-resource language contexts, revealing significant gaps.

Contribution

The paper presents BengaliFig, a novel, richly annotated dataset for assessing LLMs' reasoning in Bengali, addressing a critical low-resource and cultural reasoning gap.

Findings

LLMs show weaknesses in metaphorical reasoning

Cultural reasoning remains challenging for current models

Dataset enables diagnostic evaluation of low-resource language models

Abstract

Large language models excel on broad multilingual benchmarks but remain to be evaluated extensively in figurative and culturally grounded reasoning, especially in low-resource contexts. We present BengaliFig, a compact yet richly annotated challenge set that targets this gap in Bengali, a widely spoken low-resourced language. The dataset contains 435 unique riddles drawn from Bengali oral and literary traditions. Each item is annotated along five orthogonal dimensions capturing reasoning type, trap type, cultural depth, answer category, and difficulty, and is automatically converted to multiple-choice format through a constraint-aware, AI-assisted pipeline. We evaluate eight frontier LLMs from major providers under zero-shot and few-shot chain-of-thought prompting, revealing consistent weaknesses in metaphorical and culturally specific reasoning. BengaliFig thus contributes both a diagnostic probe for evaluating LLM robustness in low-resource cultural contexts and a step toward inclusive and heritage-aware NLP evaluation.