Metraipolis › archive › dc_cc393a4ed663

Shared Low-rank Basis Factorization for Data-free Mixture-of-Experts Compression — arXiv:2610.09342

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.09342
License (as archived)
CC0-1.0
Archived
2026-10-08T11:10:53.891Z
SHA-256 of the archived text
1e49a561317831a8f61bb89eeefed36e7b311cee70f7e60f2894aab100298f07
Stable URL of this page
https://metraipolis.com/documento/dc_cc393a4ed663

How to cite

Shared Low-rank Basis Factorization for Data-free Mixture-of-Experts Compression — arXiv:2610.09342 — arXiv (en), https://arxiv.org/abs/2610.09342 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_cc393a4ed663 (Metraipolis archive dc_cc393a4ed663, sha256 1e49a561317831a8f61bb89eeefed36e7b311cee70f7e60f2894aab100298f07)
BibTeX
@misc{dccc393a4ed663,
  title = {Shared Low-rank Basis Factorization for Data-free Mixture-of-Experts Compression — arXiv:2610.09342},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.09342},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_cc393a4ed663; sha256 1e49a561317831a8f61bb89eeefed36e7b311cee70f7e60f2894aab100298f07},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.09342. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Mixture-of-Experts (MoE) large language models decouple capacity from compute through sparse routing, but their large parameter count creates storage and serving challenges. We analyze three MoE compression families: expert pruning, expert merging, and weight reconstruction, and derive structural error bounds showing that pruning and merging can incur non-vanishing errors tied to routing and expert heterogeneity. In contrast, weight reconstruction avoids these structural costs by preserving expert structure and routing. Motivated by the analysis, we propose Shared Low-rank Basis Factorization (SLBF), a data-free weight reconstruction method that uses rank-$k$ bases shared among experts, enabling richer cross-expert sharing, faster convergence, and lower reconstruction error. A post-hoc gauge fixing removes redundant parameters at no representational cost. Across five MoE architectures spanning 16B to 122B parameters, SLBF consistently outperforms methods from all three compression families.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.09342 — 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 2026-10-08T10:14:06.077Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).