Metraipolis › archive › dc_ac6bfdaeab4e

CHASE: Channel-Aligned Structure Exploitation for Geometry-Aware Model Engineering — arXiv:2610.09476

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.09476
License (as archived): CC0-1.0
Archived: 2026-10-08T11:26:32.961Z
SHA-256 of the archived text: 12bc809f6df1d83703f80ba0a8f9324ec5d37ab82aea448c0972e31529eac53e
Stable URL of this page: https://metraipolis.com/documento/dc_ac6bfdaeab4e

How to cite

CHASE: Channel-Aligned Structure Exploitation for Geometry-Aware Model Engineering — arXiv:2610.09476 — arXiv (en), https://arxiv.org/abs/2610.09476 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ac6bfdaeab4e (Metraipolis archive dc_ac6bfdaeab4e, sha256 12bc809f6df1d83703f80ba0a8f9324ec5d37ab82aea448c0972e31529eac53e)

BibTeX

@misc{dcac6bfdaeab4e,
  title = {CHASE: Channel-Aligned Structure Exploitation for Geometry-Aware Model Engineering — arXiv:2610.09476},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.09476},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ac6bfdaeab4e; sha256 12bc809f6df1d83703f80ba0a8f9324ec5d37ab82aea448c0972e31529eac53e},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.09476. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Geometric and Spectral Alignment (GSA) characterizes trained networks through spectral concentration, physical-channel alignment, support structure, and changes in singular bases. In this paper, we propose CHASE (Channel-Aligned Structure Exploitation) to use these structures in practical model design. CHASE covers six applications across model modification, reconfiguration, and compression. CORA, COEC, and CORAM apply GSA to parameter-efficient finetuning, structured-pruning compensation, and model merging. We further develop three new methods. CAGA uses GSA to identify multi-head attention heads that can share a KV representation and constructs the shared key and value heads through geometric alignment and low-rank subspace extraction. SAKV uses GSA to determine which adjacent layers can share a low-rank KV-cache representation and the retained rank for each layer group. CAPS uses GSA spectral structure to group output neurons and selects retained input channels separately for each group. Results from CORA, COEC, and CORAM establish the effectiveness of GSA for adaptation, pruning compensation, and model merging. Experiments on CAGA show that geometric shared-head construction substantially improves MHA-to-GQA conversion, and SAKV and CAPS improve over representative baselines for KV-cache compression and structured pruning. These results show that the structures identified by GSA can be used directly to design methods for a range of model operations.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.09476 — 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 2026-10-08T10:24:06.377Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).