Test-Time Scaling of Reasoning Models for Machine Translation

Scaling Inference in Translation: framing and aims At first glance, the work interrogates whether more computation at inference — that is, letting models think longer — actually improves translation quality. The central experiment studies Test-Time S...