"AGI" team at SHROOM-CAP: Data-Centric Approach to Multilingual Hallucination Detection using XLM-RoBERTa
This work addresses the challenge of reliable AI systems for multilingual scientific hallucination detection, particularly benefiting low-resource languages, though it is incremental as it builds on existing methods with enhanced data.
The paper tackled the problem of detecting hallucinations in multilingual scientific text generated by LLMs by adopting a data-centric approach, unifying and balancing five datasets to create a training corpus of 124,821 samples, which led to competitive performance including 2nd place in Gujarati with a Factuality F1 of 0.5107.
The detection of hallucinations in multilingual scientific text generated by Large Language Models (LLMs) presents significant challenges for reliable AI systems. This paper describes our submission to the SHROOM-CAP 2025 shared task on scientific hallucination detection across 9 languages. Unlike most approaches that focus primarily on model architecture, we adopted a data-centric strategy that addressed the critical issue of training data scarcity and imbalance. We unify and balance five existing datasets to create a comprehensive training corpus of 124,821 samples (50% correct, 50% hallucinated), representing a 172x increase over the original SHROOM training data. Our approach fine-tuned XLM-RoBERTa-Large with 560 million parameters on this enhanced dataset, achieves competitive performance across all languages, including \textbf{2nd place in Gujarati} (zero-shot language) with Factuality F1 of 0.5107, and rankings between 4th-6th place across the remaining 8 languages. Our results demonstrate that systematic data curation can significantly outperform architectural innovations alone, particularly for low-resource languages in zero-shot settings.