DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads Paper • 2410.10819 • Published Oct 14, 2024 • 8
