HyPER

Dynamic test-time scaling. A closed-loop controller for multi-path LLM reasoning that reallocates compute between exploration and exploitation under a fixed budget, combining branch expansion, path reduction, and multi-granularity scheduling. PDF 路 Code

SPEX

Speculative Tree-of-Thought serving. An SGLang-based system that removes reward-dependency bottlenecks through speculative branch expansion, cross-query budget coordination, and adaptive early termination. PDF 路 Code

MAServeBench

arXiv preprint (October 2026). Composable hierarchical collaboration workloads and dependency-preserving Hierarchical Semantic Traces enable fixed-work comparisons of multi-agent serving systems, linking task completion to runtime and GPU/NPU hardware behavior. PDF 路 Code

HDA-MoE

MoE on 3D near-memory processing. Hybrid parallelism and dynamic adaptive scheduling for efficient MoE inference with memory-centric architectures. PDF 路 Code

S虏-MoE

Self-speculative MoE inference. Efficient self-speculative decoding for Mixture-of-Experts models on resource-constrained edge devices. PDF 路 Code

Tetris

HPCA 2027 (accepted). Accelerating composable test-time scaling on GPU-PIM systems via adaptive mapping, control, and memory management.