MultiKernelBench is a 285-task, 14-category benchmark for LLM-generated deep learning kernels on three hardware platforms, with evidence that category-matched one-shot prompts improve accuracy on under-represented platforms.
Pallas: a JAX kernel language
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.DC 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
MultiKernelBench: A Multi-Platform Benchmark for Kernel Generation
MultiKernelBench is a 285-task, 14-category benchmark for LLM-generated deep learning kernels on three hardware platforms, with evidence that category-matched one-shot prompts improve accuracy on under-represented platforms.