Metraipolis › archive › dc_05d7b100ca73

Trust the Direction, Search the Step: Zero-and-First-Order Methods for LLM Fine-Tuning — arXiv:2610.02190

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.02190
License (as archived)
CC0-1.0
Archived
2026-10-06T16:01:54.519Z
SHA-256 of the archived text
e918c1c5463d9cf93d37084bebc2a3417f3b327da12d340f1f2c7460c18dd06e
Stable URL of this page
https://metraipolis.com/documento/dc_05d7b100ca73

How to cite

Trust the Direction, Search the Step: Zero-and-First-Order Methods for LLM Fine-Tuning — arXiv:2610.02190 — arXiv (en), https://arxiv.org/abs/2610.02190 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_05d7b100ca73 (Metraipolis archive dc_05d7b100ca73, sha256 e918c1c5463d9cf93d37084bebc2a3417f3b327da12d340f1f2c7460c18dd06e)
BibTeX
@misc{dc05d7b100ca73,
  title = {Trust the Direction, Search the Step: Zero-and-First-Order Methods for LLM Fine-Tuning — arXiv:2610.02190},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.02190},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_05d7b100ca73; sha256 e918c1c5463d9cf93d37084bebc2a3417f3b327da12d340f1f2c7460c18dd06e},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.02190. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Step-size selection remains a central challenge in large-scale neural network optimization; conservative steps slow convergence, while aggressive steps can destabilize it. We combine \textbf{Z}ero-and-\textbf{F}irst-\textbf{O}rder optimization~(ZFO) and propose a lightweight framework that decouples direction selection from step-size. ZFO uses a trusted first-order optimizer to determine the direction and performs zeroth-order evaluations only along this one-dimensional subspace to choose how far to move. Using the current {gradient information} and two additional objective function evaluations, ZFO instances construct a local model of the objective function along the proposed direction and select a curvature-aware step within a bounded search interval. This yields an adaptive step-selection mechanism that costs less than a full line search. We provide theoretical guarantees to show that shared-sample evaluations produce reliable finite-difference curvature estimates, that the induced local model selects a near-optimal step along the search interval, and that ZFO converges to a neighborhood of a stationary point. Across the evaluated settings, language models and datasets, ZFO frequently improves optimization and final performance relative to fixed-step first-order baselines, with the magnitude and preferred local model depending on the objective. Our code is publicly available at: this https URL .

Tema degli esploratori: matematica e informatica teorica (retro-catalogo arXiv: annunci del 2026-10-02 in math.OC).
Categorie arXiv: Machine Learning (cs.LG) ; Optimization and Control (math.OC).
Pagina del lavoro: https://arxiv.org/abs/2610.02190 — 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 2026-10-06T15:14:58.016Z dalla pagina https://arxiv.org/catchup/math.OC/2026-10-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).