Jidong Zhai
Professor
Works on parallel computing, compilers, and performance evaluation, including systems for large-model training such as SmartMoE.
Jidong Zhai's lab at Tsinghua works on efficient training & inference, distributed, cloud & edge systems and interpretability, evaluation & safety of LMs. 64 papers since 2023, 14 at top venues, mostly SIGMOD and AAAI. 19 PhD students and 6 master's students. 16 alumni, most often to ByteDance and Huawei. Applications close .
At a glance
- papers, 2023–2026
- 17/ year
- at T1 venues
- 22%
- PhD students
- 19
- MS students
- 6
- alumni listed
- 16
Papers by yearfrom the lab homepage, 2023–2026
- 21 papers2026 · T1 1 · collecting
- 14 papers2025 · T1 2
- 15 papers2024 · T1 4
- 14 papers2023 · T1 7
T1 venuesOtherVenues: SIGMOD, AAAI, OSDI, EMNLP, EuroSys
Research focuspaper titles that use each topic's words
- Efficient training & inference20 papers
- Distributed, cloud & edge systems10 papers
- Interpretability, evaluation & safety of LMs0 papers
- parallel computing0 papers
- large-model training0 papers
Paper list
From the lab homepage.
202621 papers
- Frontiers Comput. SciA comprehensive taxonomy of prompt engineering techniques for large language models (opens in a new tab)
- Big Data Min. AnalA Hybrid Platform for Multi-Neurons Model with Optimum Co-Design Method (opens in a new tab)
- Tsinghua Science and TechnologyA Survey on Accelerated Technologies for Mixture-of-Experts Model Training Systems (opens in a new tab)
- Proceedings of the ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming, PPoPP , Sydney, NSW, Australia, 31 January - 4 FebruaryChituDiffusion: A Data-Characteristic-Aware Serving System for Diffusion Models (opens in a new tab)
- IEEE Trans. Parallel Distributed SystComStar: Compression-Aware Stream Query for Heterogeneous Hybrid Architecture (opens in a new tab)
- Future Gener. Comput. SystConcord: A GPU cluster scheduler with enhanced interference profiling and asymmetric job packing (opens in a new tab)
- IEEE Trans. Parallel Distributed SystCritique of \"Data Flow Lifecycles for Optimizing Workflow Coordination\" by SCC Team From Tsinghua University (opens in a new tab)
- Proceedings of the International Symposium on High-Performance Parallel and Distributed Computing, HPDC , Cleveland, OH, USA, July 13-16DiTango: Cost-Effective Parallel Diffusion Generation with Selective Attention State Reuse (opens in a new tab)
- IEEE Trans. Parallel Distributed SystEcoxPU: Autonomous Multi-Agent Orchestration for Energy-Efficient Disaggregated LLM Inference on Heterogeneous xPUs (opens in a new tab)
- IEEE Trans. Parallel Distributed SystEnabling Tile-Based Direct Query on Adaptively Compressed Data With GPU Acceleration (opens in a new tab)
- IEEE International Symposium on High Performance Computer Architecture, HPCA , Sydney, Australia, January 31 - Feb. 4FlashFuser: Expanding the Scale of Kernel Fusion for Compute-Intensive Operators via Inter-Core Connection (opens in a new tab)
- Network and Parallel Computing - IFIP WG 10.3 International Conference, NPC , Nha Trang, Vietnam, November 14-16, , Proceedings, Part IHSampler : Optimizing Multi-GPU GNN Sampling with Collision-Avoid Selection (opens in a new tab)
202514 papers
- IEEE Trans. Serv. ComputA Reinforcement Learning Based Framework for Holistic Energy Optimization of Sustainable Cloud Data Centers (opens in a new tab)
- IEEE Trans. Cloud ComputBreaking the Edge: Enabling Efficient Neural Network Inference on Integrated Edge Devices (opens in a new tab)
- Proceedings of the ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming, PPoPP , Las Vegas, NV, USA, March 1-5FlashTensor: Optimizing Tensor Programs by Leveraging Fine-grained Tensor Property (opens in a new tab)
- SIGMOD (T1)HARMONY: A Scalable Distributed Vector Database for High-Throughput Approximate Nearest Neighbor Search (opens in a new tab)
- Proceedings of the USENIX Annual Technical Conference, USENIX ATC , Boston, MA, USA, July 7-9HypeReca: Distributed Heterogeneous In-Memory Embedding Database for Training Recommender Models (opens in a new tab)
- Proceedings of the ACM/IEEE International Symposium on Code Generation and Optimization, CGO , Las Vegas, NV, USA, March 1-5IntelliGen: Instruction-Level Auto-tuning for Tensor Program with Monotonic Memory Optimization (opens in a new tab)
- IEEE Trans. Parallel Distributed SystLeveraging Graph Analysis to Pinpoint Root Causes of Scalability Issues for Parallel Applications (opens in a new tab)
- Proceedings of the USENIX Annual Technical Conference, USENIX ATC , Boston, MA, USA, July 7-9mTuner: Accelerating Parameter-Efficient Fine-Tuning on Multi-GPU Servers with Elastic Tensor (opens in a new tab)
- Proceedings of the USENIX Annual Technical Conference, USENIX ATC , Boston, MA, USA, July 7-9PluS: Highly Efficient and Expandable ML Compiler with Pluggable Graph Schedules (opens in a new tab)
- Proceedings of the USENIX Annual Technical Conference, USENIX ATC , Boston, MA, USA, July 7-9QFactory: Accelerating Quantized Large Language Model Serving with Qtile Graphs (opens in a new tab)
- Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, SC , St. Louis, MO, USA, November 16-21TraceFlow: Efficient Trace Analysis for Large-Scale Parallel Applications via Interaction Pattern-Aware Trace Distribution (opens in a new tab)
- Big Data Min. AnalTraining Large Models on Heterogeneous and Geo-Distributed Resource with Constricted Networks (opens in a new tab)
202415 papers
- IEEE Trans. Parallel Distributed SystA Survey on Performance Modeling and Prediction for Distributed DNN Training (opens in a new tab)
- TKDE (T1)Compressed Data Direct Computing for Databases (opens in a new tab)
- CCF Trans. High Perform. ComputEditorial for the special issue on programming models and system software for High-Performance Computing (HPC) environments (opens in a new tab)
- IEEE Trans. Parallel Distributed SystEfficient Inference for Pruned CNN Models on Mobile Devices With Holistic Sparsity Alignment (opens in a new tab)
- ICDE (T1)F-TADOC: FPGA-Based Text Analytics Directly on Compression with HLS (opens in a new tab)
- VLDB (T1)FreshGNN: Reducing Memory Access via Stable Historical Embeddings for Graph Neural Network Training (opens in a new tab)
- IEEE Trans. Parallel Distributed SystG-Learned Index: Enabling Efficient Learned Index on GPU (opens in a new tab)
- IEEE Trans. Parallel Distributed SystGraph-Centric Performance Analysis for Large-Scale Parallel Applications (opens in a new tab)
- Proceedings of the USENIX Annual Technical Conference, USENIX ATC , Santa Clara, CA, USA, July 10-12MAGPY: Compiling Eager Mode DNN Programs by Monitoring Execution States (opens in a new tab)
- Proceedings of the International Conference for High Performance Computing, Networking, Storage, and Analysis, SC , Atlanta, GA, USA, November 17-22MixQ: Taming Dynamic Outliers in Mixed-Precision Quantization by Online Prediction (opens in a new tab)
- Proceedings of the ACM International Conference on Architectural Support for Programming Languages and Operating SystemsOptimal Kernel Orchestration for Tensor Programs with Korch (opens in a new tab)
- IEEE Trans. Parallel Distributed SystOptimizing I/O Performance Through Effective vCPU Scheduling Interference Management (opens in a new tab)
202314 papers
- SIGMOD (T1)BladeDISC: Optimizing Dynamic Shape Machine Learning Workloads via Compiler Approach (opens in a new tab)
- OSDI (T1)Cocktailer: Analyzing and Optimizing Dynamic Control Flow in Deep Learning (opens in a new tab)
- SIGMOD (T1)CompressGraph: Efficient Parallel Graph Analytics with Rule-Based Compression (opens in a new tab)
- IEEE Trans. Parallel Distributed SystCritique of \"A Parallel Framework for Constraint-Based Bayesian Network Learning via Markov Blanket Discovery\" by SCC Team From Tsinghua University (opens in a new tab)
- OSDI (T1)EINNET: Optimizing Tensor Programs with Derivation-Based Transformations (opens in a new tab)
- IEEE Trans. Parallel Distributed SystEnabling Efficient Random Access to Hierarchically Compressed Text Data on Diverse GPU Platforms (opens in a new tab)
- ICLR (T1)GLM-130B: An Open Bilingual Pre-trained Model (opens in a new tab)
- Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, SC , Denver, CO, USA, November 12-17GraphSet: High Performance Graph Mining through Equivalent Set Transformations (opens in a new tab)
- EMNLP (T1)Joint Geometrical and Statistical Domain Adaptation for Cross-domain Code Vulnerability Detection (opens in a new tab)
- IEEE Trans. ComputersOptimizing DNNs With Partially Equivalent Transformations and Automated Corrections (opens in a new tab)
- Proceedings of the USENIX Annual Technical Conference, USENIX ATC , Boston, MA, USA, July 10-12SmartMoE: Efficiently Training Sparsely-Activated Models through Combining Offline and Online Parallelization (opens in a new tab)
- Softw. Pract. ExpSpecial issue on new trends in high-performance computing: Software systems and applications (opens in a new tab)
Where alumni went16 listed, 62% to industry
Industry 8Academia 5
Most often to
- ByteDance
- Huawei
- Meta
Also to
- Tsinghua University
- Carnegie Mellon University
- UT Austin
Apply to TsinghuaFall 2027
Computer Science PhD (international applicants)
- Round
- Main
- Opens
- GRE
- See program page
- English
- Per school catalog
- Fee
- RMB 800
Not yet confirmed for this intake.
Similar labsnearest by research summary
- USUT Austin
José del R. Millán
- Interfaces & interaction techniques
- CNPeking
Xin Jin
- Efficient training & inference
- USUCSD
Yiying Zhang
- Efficient training & inference
- CNSJTU
Chaoyue Niu
- Distributed, cloud & edge systems
- USUCSD
Mingu Kang
- Accelerators, chips & architecture
- CNSJTU
Tao Song
- Distributed, cloud & edge systems
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- The lab homepage
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.
Tsinghuacloses