Touseef Hasan

Bigger isn't always better: smaller, task-aligned LLMs can exhibit more consistent annotation behavior than their larger counterparts in sensitive, low-resource settings like Bangla hate speech detection.