---
format: "aidr-story-markdown/v1"
id: "67626b0f879eccd9285823e3b7de72256b2bff259702c70656951b26898c34f8"
canonical_url: "https://aidr.today/67626b0f?lang=en"
title: "KernelMosaic: An Open Dataset and Agentic Framework for CUDA Kernel Generation"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-11-13T05:00:00.000Z"
category: "Data"
topics: ["agent","data-engineering"]
source_urls: ["https://arxiv.org/abs/2611.14604"]
summary: "KernelMosaic is an open dataset pairing diverse CUDA tasks with verified solutions and trajectories. An agentic framework guides a strong LLM through implementation, correctness repair, and performance optimization. Supervised fine-tuning of a small open-source model on KernelMosaic raised correctness on KernelBench L1, L2, and L3 to 88%, 82%, and 16% respectively."
---

# KernelMosaic: An Open Dataset and Agentic Framework for CUDA Kernel Generation

> [Open the canonical story](<https://aidr.today/67626b0f?lang=en>)

**Published:** 2026-11-13T05:00:00.000Z
**Category:** Data
**Topics:** agent, data\-engineering

## Summary

KernelMosaic is an open dataset pairing diverse CUDA tasks with verified solutions and trajectories\. An agentic framework guides a strong LLM through implementation, correctness repair, and performance optimization\. Supervised fine\-tuning of a small open\-source model on KernelMosaic raised correctness on KernelBench L1, L2, and L3 to 88%, 82%, and 16% respectively\.

## Sources

- [Story source](<https://arxiv.org/abs/2611.14604>)