Versatile On-device Adaptation at the Edge by Unifying Few-shot, Zero-shot, Continual, and In-context Learning

2026-07-31

AI Summary

This work addresses the challenge of enabling unified, real-time, and privacy-preserving online learning on resource-constrained edge devices. The authors propose Embedded-Centric Learning (ECL), a novel framework that, for the first time, integrates few-shot, zero-shot, continual, and in-context learning paradigms into a single hardware-deployable system, eliminating reliance on cloud infrastructure. Designed for low-power edge chips, ECL achieves 96.8% accuracy on Omniglot under a 5-way 1-shot setting, establishes the first hardware baseline for keyword continual learning at 71.8%, and demonstrates the first on-device hardware validation of zero-shot (60.6%) and in-context learning (46.2%) at the edge.

Abstract

With the ever-increasing pervasiveness of smart edge devices, the demand is growing for applications that can be tailored to users (e.g., custom keyword spotting) or patients (e.g., adaptive health monitoring). Yet, most edge devices rely on fixed inference algorithms and thus cannot learn on-device to personalize predictions. When they can, devices typically support only a specific learning scenario, such as few-shot learning (FSL): going beyond this requires resorting either to another specialized device or to cloud-based retraining, which implies significant energy and latency overheads, a lack of real-time capabilities, and privacy concerns. In this work, we introduce embedder-centric learning (ECL), a framework that unifies four different online learning scenarios: FSL for on-the-fly customization, continual learning (CL) for knowledge accumulation, zero-shot learning (ZSL) for leveraging semantic data, and in-context learning (ICL) for adapting beyond classification. We demonstrate in silicon that ECL can be deployed on resource-constrained devices across four real-world use cases representative of the aforementioned learning scenarios. Our approach establishes a new state-of-the-art performance for FSL character recognition (Omniglot: 96.8% for 5-way 1-shot, 83.3% for 32-way 1-shot), and the first hardware baseline for CL in keyword spotting (NeuroBench keyword FSCIL: 71.8% for 200-way 5-shot). Moreover, we present the first hardware demonstrations of ZSL with semantic data (60.6% for 5-way spoken sentence classification) and ICL (46.2% at the 500th token of RegBench) operating at micro-to-milliwatt power budgets. Therefore, by unifying multiple learning scenarios, we pave the way for smart and versatile devices that can adapt right at the edge, without reliance on the cloud.

Problem

• on-device learning
• edge computing
• few-shot learning
• continual learning
• zero-shot learning

Innovation

• embedder-centric learning
• on-device adaptation
• few-shot learning
• continual learning
• edge AI