Metraipolis › archive › dc_ce0e33fa4478

CANNBot-CAKE: Collaborative Agentic Kernel Evolution for Ascend NPU — arXiv:2610.31034

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.31034
License (as archived)
CC0-1.0
Archived
2026-10-25T20:53:12.101Z
SHA-256 of the archived text
94eb6a03872934b82d7b706819218a4a5f1c8bde8e85ae5376eba95d5be5102a
Stable URL of this page
https://metraipolis.com/documento/dc_ce0e33fa4478

How to cite

CANNBot-CAKE: Collaborative Agentic Kernel Evolution for Ascend NPU — arXiv:2610.31034 — arXiv (en), https://arxiv.org/abs/2610.31034 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ce0e33fa4478 (Metraipolis archive dc_ce0e33fa4478, sha256 94eb6a03872934b82d7b706819218a4a5f1c8bde8e85ae5376eba95d5be5102a)
BibTeX
@misc{dcce0e33fa4478,
  title = {CANNBot-CAKE: Collaborative Agentic Kernel Evolution for Ascend NPU — arXiv:2610.31034},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.31034},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ce0e33fa4478; sha256 94eb6a03872934b82d7b706819218a4a5f1c8bde8e85ae5376eba95d5be5102a},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.31034. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Efficient custom operators are essential for modern deep learning workloads, yet developing high-performance kernels for Ascend NPUs remains slow, expert-intensive, and difficult to automate. Although AI agents now approach expert performance on GPU kernel benchmarks, current models remain less capable of generating AscendC kernels for NPUs, potentially reflecting the relative scarcity of relevant training data. We thus present CAKE, a CANN-native system for deployable custom operators that combines adaptive multi-DSL routing, separate Generation and Evolution modes, and reviewed knowledge reuse through CAKE Book and a Self-Evolving Agent. On 53 CANN-Bench v1.1 operators on Ascend 950PR, CAKE's aggregated results achieve $100\%$ case pass rate on the standard set (1,060/1,060) and $98.28\%$ on the hidden set (4,167/4,240), with an overall mean score of 78.63 out of 100 and an average performance speedup of $1.334\times$ over the baseline. Evolution then pushes correct operators further: EngramGateFusion cuts aggregate device time by $46.6\%$ over 20 rounds and roughly halves the mean roofline gap. These results offer a practical path toward closing Ascend's AI-for-kernel capability gap.

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/2610.31034 — 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 2026-10-25T20:47:56.894Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).