Nov 2026· Proceedings of the 34th ACM International Conference on Multimedia· 0 citations· 6 references
TL;DR
SkillCache is proposed, which caches model capabilities via multi-view coupled skill operators for low-latency multimodal edge intelligence and enables the discovery, routing, maintenance, and personalization of capabilities through simple coefficient operations.
Abstract
Multimodal assistants are increasingly deployed on resource-constrained devices, yet the core trade-off remains unresolved: Large Language Models (LLMs) are too heavy for on-device execution, while full cloud offloading incurs high latency, bandwidth costs, and privacy risks. Existing edge-cloud collaborative systems rely on self-evaluative routing (SER), where the edge model decides whether to respond locally or offload to the cloud. On one hand, we find this strategy unreliable due to the inherent capability limitations of small models on edge devices. On the other hand, real-world model usage exhibits clustering, temporal stability, and partially shared topic characteristics across users. Based on these observations, we argue that the edge should act not as a general assistant, but as a cache of frequently used capabilities. In this paper, we propose SkillCache, which caches model capabilities via multi-view coupled skill operators for low-latency multimodal edge intelligence. In SkillCache, capabilities are represented in a unified space and can be combined through sparse composition over shared multimodal bases. As a result, SkillCache enables the discovery, routing, maintenance, and personalization of capabilities through simple coefficient operations. We implement a prototype of SkillCache using typical edge and cloud models, and extensive experimental results show that SkillCache closes 62.0% of the edge–cloud quality gap (Core GC) on MMBench, OCRBench v2, and CharXiv at the matched 34% cloud budget, with smaller gains on the harder BLINK and MMMU stress tests. It reduces uplink traffic by 88% (85.34% net bidirectional reduction after amortizing skill-pack delivery) with only 30 ms routing overhead.
Supporting data, adapters, predictions and code for the article *Low-Cost LoRA Fine-Tuning of Small Language Models for Multi-Step Arithmetic Reasoning* by Jake O'Grady, Asena Isik Gürhan, Chee Fong Ting and Effirul Ramlan (University of Galway). We generated 20,000 GSM8K-derived arithmetic problems with step-by-step s...
O'Grady, Jake, Gürhan, Asena Isik, Chee, Fong Ting et al.· Zenodo (CERN European Organi...· 465 citations
The results are packaged in the Greenfield Startup Model (GSM), which explains the priority of startups to release the product as quickly as possible, and the need to shorten time-to-market, by speeding up the development through low-precision engineering activities.
Carmine Giardino, Nicolò Paternoster, M. Unterkalmsteiner et al.· IEEE Transactions on Softwar...· 178 citations· ⚡14
Software startup companies develop innovative, software-intensive products within limited timeframes and with few resources, searching for sustainable and scalable business models.
M. Unterkalmsteiner, P. Abrahamsson, Xiaofeng Wang et al.· e-Informatica Software Engin...· 157 citations· ⚡17
This study conducts a case survey study based on the secondary data of the major pivots happened in 49 software startups, and demonstrates that customer need pivot is the most common among all pivot types.
Sohaib Shahid Bajwa, Xiaofeng Wang, Anh Nguyen-Duc et al.· Empirical Software Engineeri...· 127 citations· ⚡15
The comparison of adopter and non-adopter sample reveals three potential adoption inhibitor, security, data privacy, and portability, which underlines the importance of the technical and security perspectives for research investigating the adoption of technology.
Nattakarn Phaphoom, Xiaofeng Wang, S. Samuel et al.· Journal of Systems and Softw...· 111 citations· ⚡8
Objective There is dearth of information on the timeliness of antenatal care (ANC) uptake. This study aimed to determine the timely ANC uptake by a medically trained provider (MTP) as per the World Health Organization (WHO) recommendations and the country guideline. Methods Cross-sectional survey was done with 2,731 wo...
B. Sarker, Musfikur Rahman, T. Rahman et al.· PLoS ONE· 89 citations· ⚡7
By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.