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 updateTalks, papers, tools, and moments from CAT Lab.
HPC-Bench, our benchmark for evaluating LLM-generated HPC code optimizations, has been accepted in the NeurIPS 2026 Evaluations and Datasets track.
Read updateOur accelerator-free NKI teaching toolchain has been accepted at SIGCSE TS 2027.
Read updateOur collaboration on cross-scenario generality of agent memory systems and the AutoMEM baseline has been accepted at AACL-IJCNLP 2026.
Read updateTileBench, our benchmark for performance evaluation and bottleneck diagnosis of tile-based programming models, has been accepted at AACL-IJCNLP 2026.
Read updateOur paper LLMPROF, on top-down performance profiling for LLM serving systems, has been accepted at SC 2026.
Read updateKeren Zhou presented Proton’s latest progress toward very low-overhead, always-on profiling for GPU-accelerated LLM workloads.
Read updateThai-Hoa presented TileLens at the May 2026 Triton community meetup.
Read updateKeren (along with Peter Bell and Mario Lezcano from OpenAI) gave a GPU MODE talk on Gluon and linear layouts, including a demo of Thai-Hoa’s visualizer.
Read updateOur paper on context-aware feedback compression for online judge programming has been accepted at FSE-IVR 2026.
Read updateOur paper Triton-Sanitizer has been accepted at ASPLOS 2026. Congratulations to all the authors!
Read update