Metraipolis › archive › dc_f1053a02e1ff
Beyond Experience: Knowledge-Guided Evolution for NPU Kernel Generation — arXiv:2610.27588
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.27588
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-21T07:06:40.767Z
- SHA-256 of the archived text
016ff7b6bc5de9a18564a7f580448a22dd3dd7b6a9466c2ae89ebc6d8925bc95- Stable URL of this page
https://metraipolis.com/documento/dc_f1053a02e1ff
How to cite
Beyond Experience: Knowledge-Guided Evolution for NPU Kernel Generation — arXiv:2610.27588 — arXiv (en), https://arxiv.org/abs/2610.27588 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f1053a02e1ff (Metraipolis archive dc_f1053a02e1ff, sha256 016ff7b6bc5de9a18564a7f580448a22dd3dd7b6a9466c2ae89ebc6d8925bc95)
BibTeX
@misc{dcf1053a02e1ff,
title = {Beyond Experience: Knowledge-Guided Evolution for NPU Kernel Generation — arXiv:2610.27588},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.27588},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f1053a02e1ff; sha256 016ff7b6bc5de9a18564a7f580448a22dd3dd7b6a9466c2ae89ebc6d8925bc95},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.27588. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
LLM agents have become capable kernel engineers on mainstream GPU stacks, but they struggle on emerging accelerators such as Ascend NPUs, whose programming model is barely represented in pretraining data. Recent NPU agents address this gap largely through experience: they accumulate memory from their own generation and optimization trials and reuse it on later tasks. We argue that experience alone is not enough for kernel generation. An agent can only remember what it has tried, and it rarely tries a mechanism it does not know. As a result, such memory offers little help when a new operator calls for a different mechanism, and even familiar operators raise unseen situations as optimization proceeds. We present ASCENDEVOLVE, which grounds every step of an evolutionary kernel search in a large external knowledge base built from platform documentation, toolkit headers, and operator repositories. Because this corpus far exceeds any context window, ASCENDEVOLVE indexes it and routes each agent to a small, task-specific slice of evidence instead of the corpus itself. The retrieved knowledge seeds diverse implementation strategies, guides repair when a candidate fails, and constrains how strong candidates are recombined. On 53 Ascend C operators from CANN-Bench spanning four difficulty levels, ASCENDEVOLVE achieves an overall score of 71.39, improving over Reflexion and OpenEvolve by 11.6% and 17.2%. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.27588 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-21T07:05:43.424Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).