analiza aiqbench
This arXiv paper proposes a transfer learning method that trains a diffusion model over parameter-efficient fine-tuning (PEFT) weights from upstream tasks, then selects a downstream adapter from a finite set of sampled candidates. Restricting the hypothesis space to a finite set is what lets the authors derive tighter risk bounds than standard continuous-weight approaches, which tend to yield vacuous bounds in the low-data regime. Readers should note that the claim is validated on bounds and comparisons in low-shot settings, so practical value depends on whether the guarantees hold up at scale and across more diverse tasks.
Weight Space Learning for Certifiable Few-shot Transfer Learning
arXiv:2502.06970v3 Announce Type: replace Abstract: In contemporary deep learning, a prevalent and effective workflow for solving low-data problems is adapting powerful pre-trained foundation models (FMs) to new tasks via parameter-efficient fine-tuning (PEFT). However, while empirically effective, the resulting solutions lack generalisation guarantees to certify their accuracy - which may be required for ethical or legal reasons prior to deployment in high-importance applications. In this paper we develop a novel transfer learning approach that is designed to facilitate non-vacuous learning theoretic generalisation guarantees for downstream tasks, even in the low-shot regime. Specifically, we first use upstream tasks to train a distribution over PEFT parameters. We then learn the downstream task by a sample-and-evaluate procedure -- sampling plausible PEFTs from the trained diffusion model and selecting the one with the highest likelihood on the downstream data. Crucially, this confines our model hypothesis to a finite set of PEFT samples. In contrast to the typical continuous hypothesis spaces of neural network weights, this facilitates tighter risk certificates. We instantiate our bound and show non-trivial generalization guarantees compared to existing learning approaches which lead to vacuous bounds in the low-shot regime.
Abstract:In contemporary deep learning, a prevalent and effective workflow for solving low-data problems is adapting powerful pre-trained foundation models (FMs) to new tasks via parameter-efficient fine-tuning (PEFT). However, while empirically effective, the resulting solutions lack generalisation guarantees to certify their accuracy - which may be required for ethical or legal reasons prior to deployment in high-importance applications. In this paper we develop a novel transfer learning approach that is designed to facilitate non-vacuous learning theoretic generalisation guarantees for downstream tasks, even in the low-shot regime. Specifically, we first use upstream tasks to train a distribution over PEFT parameters. We then learn the downstream task by a sample-and-evaluate procedure -- sampling plausible PEFTs from the trained diffusion model and selecting the one with the highest likelihood on the downstream data. Crucially, this confines our model hypothesis to a finite set of PEFT samples. In contrast to the typical continuous hypothesis spaces of neural network weights, this facilitates tighter risk certificates. We instantiate our bound and show non-trivial generalization guarantees compared to existing learning approaches which lead to vacuous bounds in the low-shot regime.
Machine Learning (cs.LG); Machine Learning (stat.ML)
arXiv:2502.06970 [cs.LG]
(or arXiv:2502.06970v3 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2502.06970 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 277 cuvinte
Subiecte conexe
Certifiable Safe Model-Based Reinforcement Learning with Control-Affine Dynamics Approximation
arXiv · 21 iul. 2026
SHIFT-JEV: Certified Abstention and Gain-Gated Continual Adaptation for LLM-Based Decision Models under Distribution Shift
arXiv · 27 oct. 2026
Nous: Learning and Certifying Memory Decisions Before Source Calibration
arXiv · 2 oct. 2026
How Accurate Is Accurate Enough?
arXiv · 1 oct. 2026