HPC-Bench Accepted at NeurIPS 2026
HPC-Bench, our benchmark for evaluating LLM-generated HPC code optimizations, has been accepted in the NeurIPS 2026 Evaluations and Datasets track.
Read updateCAT Lab at George Mason University develops compilers, architectures, and tools that make AI systems easier to understand, faster to run, and more reliable at scale.
Explore Software View Publications
Compiler · Architecture · Toolkit
Latest News
What we are publishing, presenting, and building.
NKI Teaching Lab Accepted at SIGCSE TS 2027
Our accelerator-free NKI teaching toolchain has been accepted at SIGCSE TS 2027.
Read updateAutoMEM Accepted at AACL 2026
Our collaboration on cross-scenario generality of agent memory systems and the AutoMEM baseline has been accepted at AACL-IJCNLP 2026.
Read updateFeatured Publications
Recent research from across the computing stack.
NKI Teaching Lab: An Accelerator-Free Toolchain for Learning Tile-Level AI Accelerator Programming
An accelerator-free teaching toolchain for learning tile-level AI accelerator programming.
HPC-Bench: A Comprehensive Benchmark for High Performance Computing Codes
A benchmark for evaluating LLM-generated optimizations of HPC kernels with program-level correctness checks and performance measurements.
LLMPROF: Identifying Performance Bottlenecks in LLM Serving Systems with Top-Down Profiling
Exploring Cross-Scenario Generality of Agentic Memory Systems: Diagnostics and a Strong Baseline
A cross-scenario evaluation of agent memory systems and AutoMEM, a baseline that lets agents manage memory through tools.