1AutoKernel: Autonomous GPU Kernel Optimization via Iterative Agent-Driven Search (opens in new tab)(arxiv.org)4OsamaJaber1d ago0
2DeepSeek V4's indexer OOMs at 65K context. We got it to 1M in 6G (opens in new tab)(arxiv.org)8OsamaJaber5d ago0
3Ouroboros: Dynamic Weight Generation for Recursive Transformers (opens in new tab)(arxiv.org)2OsamaJaber15d ago0
4Tide: Token-Informed Depth Execution for Per-Token Early Exit in LLM Inference (opens in new tab)(arxiv.org)3OsamaJaber21d ago1
