Minyi Guo

3 papers Shanghai Jiao Tong University

• ClusterFusion: Expanding Operator Fusion Scope for LLM Inference via Cluster-Level Collective Primitive
• Communication-Efficient Diffusion Denoising Parallelization via Reuse-then-Predict Mechanism
• Yggdrasil: Bridging Dynamic Speculation and Static Runtime for Latency-Optimal Tree-Based LLM Decoding