TokenSpeed (Repo Collaborator)

A speed-of-light LLM inference engine for agentic workloads, with TensorRT-LLM-level performance and vLLM-level usability. Features a local-SPMD modeling layer with a static parallelism compiler, a C++/Python finite-state-machine scheduler with type-safe KV-cache reuse, and a pluggable kernel system including one of the fastest MLA implementations on Blackwell.

[code]

ActionEngine: From Reactive to Programmatic GUI Agents via State Machine Memory

Hongbin Zhong, Fazle Faisal, Luis França, Tanakorn Leesatapornwongsa, Adriana Szekeres, Kexin Rong, Suman Nath

submitted to SOSP

StreamFlow: Theory, Algorithm, and Implementation for High-Efficiency Rectified Flow Generation

Sen Fang*, Hongbin Zhong*, Yalin Feng, Dimitris N. Metaxas

ICML 2026

Stable Signer: Hierarchical Sign Language Generative Model

Sen Fang*, Yalin Feng*, Hongbin Zhong, Yanxin Zhang, Dimitris N. Metaxas

ACL 2026 🏆 Best Paper Nomination

HoneyBee: Efficient Role-based Access Control for Vector Databases via Dynamic Partitioning

Hongbin Zhong, Matthew Lentz, Nina Narodytska, Adriana Szekeres, Kexin Rong

SIGMOD 2026

Fast Hypothetical Updates Evaluation

Haneen Mohammed*, Alexander Yao*, Charlie Summers*, Hongbin Zhong, Gromit Yeuk-Yin Chan, Subrata Mitra, Lampros Flokas, Eugene Wu

SIGMOD 2025 (Demo)

FaDE: More Than a Million What-ifs Per Second

Haneen Mohammed*, Alexander Yao*, Charlie Summers*, Hongbin Zhong, Gromit Yeuk-Yin Chan, Subrata Mitra, Lampros Flokas, Eugene Wu

VLDB 2025 (Implemented as a DuckDB extension)

Accelerating Deletion Interventions on OLAP Workload

Haneen Mohammed, Alexander Yao, Lampros Flokas, Hongbin Zhong, Charlie Summers, Eugene Wu

ICDE 2024

PECJ: Stream Window Join on Disorder Data Streams with Proactive Error Compensation

Xianzhi Zeng*, Shuhao Zhang, Hongbin Zhong, Hao Zhang, Mian Lu, Zhao Zheng, Yuqiang Chen

SIGMOD 2024 (Adopted by OpenMLDB)