Fajri Koto

The paper introduces IndoBias, a dual-track, culturally-grounded benchmark to evaluate biases in LLMs across Indonesian and three local languages, revealing significant differences in bias patterns across languages and data sources.

The paper shows that safety failures in low-resource languages are due to a failure in the model's safety decision calibration, not a lack of underlying knowledge, and proposes a recalibration method to fix this.

The paper introduces MIDI, a novel multilingual dataset that embeds idioms in realistic sentence and conversational contexts across diverse resource levels, revealing that idiom comprehension is significantly harder in low-resource languages and that literal interpretations pose a greater challenge than figurative ones.