<ul><li>Reliability of LLMs is questionable even as they get better at more tasks.</li><li>This work focuses on utilizing the multilingual knowledge of an LLM to inform its decision to abstain or answer when prompted.</li><li>A multilingual pipeline is developed to calibrate the model's confidence and let it abstain when uncertain.</li><li>Results show significant improvement in accuracy, with a $71.2% improvement for Bengali and a 15.5% improvement for English.</li></ul>

MKA: Leveraging Cross-Lingual Consensus for Model Abstention

Discover more