Metraipolis › archive › dc_7e8e33488e90

Scalable Kronecker-Fisher Approximation: Efficient Hessian Analysis for Billion-Parameter Language Models Compression — arXiv:2609.02451

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.02451
License (as archived)
CC0-1.0
Archived
2026-10-17T03:38:08.552Z
SHA-256 of the archived text
2d6b81da5aa9655fdc00ded7584aaf859d765e00702895c03f4a41437338a5de
Stable URL of this page
https://metraipolis.com/documento/dc_7e8e33488e90

How to cite

Scalable Kronecker-Fisher Approximation: Efficient Hessian Analysis for Billion-Parameter Language Models Compression — arXiv:2609.02451 — arXiv (en), https://arxiv.org/abs/2609.02451 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_7e8e33488e90 (Metraipolis archive dc_7e8e33488e90, sha256 2d6b81da5aa9655fdc00ded7584aaf859d765e00702895c03f4a41437338a5de)
BibTeX
@misc{dc7e8e33488e90,
  title = {Scalable Kronecker-Fisher Approximation: Efficient Hessian Analysis for Billion-Parameter Language Models Compression — arXiv:2609.02451},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.02451},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_7e8e33488e90; sha256 2d6b81da5aa9655fdc00ded7584aaf859d765e00702895c03f4a41437338a5de},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.02451. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

In this paper, we propose a scalable Kronecker-based approximation that captures cross-layer interactions without storing the entire Fisher matrix, enabling practical Hessian analysis for billion-parameter networks where full computation is infeasible. Our approach reveals consistent vulnerability patterns: value projection layers exhibit the highest sensitivity and strongest cross-layer correlations across multiple model families, while other components exhibit architecture-specific behaviors. Through extensive experiments on quantization, sparsification, inter-layer corruption, and post-corruption fine-tuning, we demonstrate that our approximation strongly correlates with both performance degradation and recovery. Our framework provides a practical, theoretically grounded tool for identifying fragile components in large models, opening new avenues for guided compression and optimization strategies, such as mixed-precision allocation, layer-wise sparsity, and adaptive low-rank decomposition across layers and even individual weight groups.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-03 in cs.AI).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.02451 — 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 2026-10-17T03:13:12.550Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-03?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).