Utkarsh Saxena
2 papers · Latest:
Natural Language Processing
Long-Context Aware Upcycling: A New Frontier for Hybrid LLM Scaling
HyLo upcycles pretrained Transformers into hybrid LLMs, extending context by 32x and reducing KV-cache memory by over 90% for efficient long-context processing.
2604.24715
Machine LearningOASIS: Online Activation Subspace Learning for Memory-Efficient Training
OASIS is an online activation subspace learning algorithm that significantly reduces memory requirements for training large language models while maintaining performance.
2604.09406
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.