AI Paper Atlas

Research map / Large language models / Agents, code and math

QiMeng-Kernel: Macro-Thinking Micro-Coding Paradigm for LLM-Based High-Performance GPU Kernel Generation

Xinguo Zhu, Shaohui Peng, Jiaming Guo, Yunji Chen, Qi Guo, Yuanbo Wen, Hang Qin, Ruizhi Chen, Qirui Zhou, Ke Gao, Yanjun Wu, Chen Zhao and 1 more

AAAI 2026 · Large language models › Agents, code and math · approach: LLMs for optimization and heuristics · 16 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add MTMC macro-thinking/micro-codingGemini 2.5 Pro · KernelBench Level 1 (H100)execute accuracy (%): 63 -> 100Improved
add MTMCGemini 2.5 Pro · KernelBench Level 3 (H100)mean speedup: 0.27 -> 0.77Improved
add MTMCGemini 2.5 Pro · KernelBench Level 3 (H100)execute accuracy (%): 36 -> 70Improved
add MTMCGemini 2.5 Flash · TritonBench-T (A100)execute accuracy (%): 9.04 -> 54.82Improved
add MTMCGemini 2.5 Flash · TritonBench-G (A100)fast1 (%): 4.35 -> 9.78Improved
add MTMCGemini 2.5 Flash · TritonBench-G (A100)fast2 (%): 1.63 -> 1.63No clear effect
generate stepwise with Micro Coding instead of single pass (ablation)Gemini 2.5 Flash · KernelBench Level 1execute accuracy (%): 60% -> 94%Improved
generate stepwise instead of single pass (ablation)Gemini 2.5 Flash · KernelBench Level 3execute accuracy (%): 10% -> 64%Improved
use RL-trained policy (DeepSeek-Coder-1.3B) instead of GPT-4o for macro thinking (ablation)MTMC · KernelBench Level 3execute accuracy (%): 40% -> 100%Improved
use hardware-aware action space for GPT-4o macro thinking (ablation)GPT-4o without policy · KernelBench Level 3execute accuracy (%): 20% -> 40%Improved
use Llama-3.2-1B instead of DeepSeek-Coder-1.3B as policy (ablation)MTMC · KernelBench Level 3execute accuracy (%): 100% -> 80%Hurt
generate CUDA instead of Triton for matmul kernels (ablation)MTMC · KernelBench matmul tasksexecution time (ms): task 6: 4.43 -> 1.34; task 2: 1.36 -> 1.66Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy