Sara Rajaee

Smaller models can outperform larger counterparts in linguistic reasoning tasks, challenging assumptions about model size and capability.

Evaluators using cESA can achieve more reliable translation quality assessments while cutting annotation time by leveraging shared context across multiple outputs.

Forget brute-force scaling: Tiny Aya proves a 3B parameter model can achieve state-of-the-art multilingual performance with clever training and region-aware specialization.

Generic reasoning hurts machine translation, but a new structured reasoning approach—with iterative drafting, refinement, and revision—unlocks significant gains.