Publications
* Equal contribution.
2026
-
EdgeKernelBench: Unified LLM-Driven Kernel Generation for Edge Backends
-
MetaKernelBench: Measuring GPU Kernel Knowledge Transfer Beyond Code
-
StreamingTOM: Streaming Token Compression for Efficient Video Understanding
-
LVOmniBench: Pioneering Long Audio-Video Understanding Evaluation for Omnimodal LLMs
-
MobileKernelBench: Can LLMs Write Efficient Kernels for Mobile Devices?
-
DICE: Diffusion Large Language Models Excel at Generating CUDA Kernels