Oussama Elachqar

ButterflyQuant: Ultra-low-bit LLM Quantization through Learnable Orthogonal Butterfly Transforms

MARVIS: Modality Adaptive Reasoning over VISualizations

When Judgment Becomes Noise: How Design Failures in LLM Judge Benchmarks Silently Undermine Validity

Do LLMs ``know'' internally when they follow instructions?