Mehwish Fatima

Riazi-8B outperforms existing Urdu models in mathematical reasoning, proving that targeted language adaptation can bridge the gap in low-resource AI applications.

Adaptive verification can drastically cut down on the cost of ensuring factual accuracy in long-form generation by focusing on the riskiest claims.

ROMEVA reveals that prioritizing embedding stability can compromise downstream performance, challenging conventional wisdom in low-resource NLP.

Textual semantics excel at predicting current affect but fail to capture future changes, which are better modeled by numeric trajectories.