Metraipolis › archive › dc_d9e014bbc681

DPA-MoE: A Runtime Precision Control Framework for MoE Inference — arXiv:2610.21378

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.21378
License (as archived)
CC0-1.0
Archived
2026-10-16T09:57:41.559Z
SHA-256 of the archived text
238804ff4721967deeb61d37e3e11c268dd0f89fd6f1fa87e7ee3e52dc423fbe
Stable URL of this page
https://metraipolis.com/documento/dc_d9e014bbc681

How to cite

DPA-MoE: A Runtime Precision Control Framework for MoE Inference — arXiv:2610.21378 — arXiv (en), https://arxiv.org/abs/2610.21378 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d9e014bbc681 (Metraipolis archive dc_d9e014bbc681, sha256 238804ff4721967deeb61d37e3e11c268dd0f89fd6f1fa87e7ee3e52dc423fbe)
BibTeX
@misc{dcd9e014bbc681,
  title = {DPA-MoE: A Runtime Precision Control Framework for MoE Inference — arXiv:2610.21378},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.21378},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d9e014bbc681; sha256 238804ff4721967deeb61d37e3e11c268dd0f89fd6f1fa87e7ee3e52dc423fbe},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.21378. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Mixture-of-Experts (MoE) models reduce computation by routing each token to a small set of experts, but each selected expert is still used at a fixed precision. We present DPA-MoE, a runtime precision control framework that makes routed experts any-precision. DPA-MoE encodes expert feed-forward weights as prefix-decodable chunks using SharedScale AnyBCQ, profiles per-expert reconstruction residuals offline, and allocates chunks at inference time using live router gate scores. This post-routing allocation lets a single encoded MoE checkpoint operate at multiple effective weight budgets without retraining or re-encoding. Experiments on Qwen3-30B-A3B and DeepSeekMoE-16B show that DPA-MoE improves KL divergence and downstream accuracy over uniform chunk allocation at matched budgets, with modest steady-state throughput overhead in our prototype.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.21378 — 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 2026-10-16T08:44:47.645Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).