Hasan Abed Al Kader Hammoud

Turning the Spell Around: Lightweight Alignment Amplification via Rank-One Safety Injection

Train Long, Think Short: Curriculum Learning for Efficient Reasoning

Beyond the Last Answer: Your Reasoning Trace Uncovers More than You Think

Reasoning Vectors: Transferring Chain-of-Thought Capabilities via Task Arithmetic

RAC-LoRA: A Theoretical Optimization Framework for Low-Rank Adaptation

SynthCLIP: Are We Ready for a Fully Synthetic CLIP Training?