FlashMLA: Efficient Multi-head Latent Attention Kernels

[MLSys'25] QServe: W4A8KV4 Quantization and System Co-design for Efficient LLM Serving; [MLSys'25] LServe: Efficient Long-sequence LLM Serving with Unified Sparse Attention

[ACL 2025] RetroLLM: Empowering LLMs to Retrieve Fine-grained Evidence within Generation