Extractive Summarization on a CMOS Ising Machine
IEEE Access · 2026 · 14 · 56724-56735
DOI: 10.1109/access.2026.3681908 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Ziqing Zeng und 6 weiteren Autor:innen. Die Arbeit nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Science Foundation, Defense Advanced Research Projects Agency (DARPA) Quantum-Inspired Classical Computing (QuICC) Program through Air Force Research Laboratory.
Abstract (Originalsprache)
Extractive summarization (ES) aims to generate a concise summary by selecting a subset of sentences from a document while maximizing relevance and minimizing redundancy. Although modern ES systems achieve high accuracy using powerful neural models, their deployment typically relies on CPU or GPU infrastructures that are energy-intensive and poorly suited for real-time inference in resource-constrained environments. In this work, we explore the feasibility of implementing McDonald-style extractive summarization on a low-power CMOS coupled oscillator-based Ising machine (COBI) that supports integer-valued, all-to-all spin couplings. We first propose a hardware-aware Ising formulation that reduces the scale imbalance between local fields and coupling terms, thereby improving robustness to coefficient quantization: this method can be applied to any problem formulation that requires $k$ of $n$ variables to be chosen. We then develop a complete ES pipeline including 1) stochastic rounding and iterative refinement to compensate for precision loss, and 2) a decomposition strategy that partitions a large ES problem into smaller Ising subproblems that can be efficiently solved on COBI and later combined. Experimental results on the CNN/DailyMail dataset show that our pipeline can produce high-quality summaries using only integer-coupled Ising hardware with limited precision. COBI achieves 3.1– $4.3\times $ speedups in runtime compared to a brute-force baseline, with performance comparable to software-based Tabu search. In terms of energy, COBI delivers three orders of magnitude reduction compared to the brute-force approach and 2.5 orders of magnitude reduction compared to Tabu search, maintaining competitive summary quality. These results highlight the potential of deploying CMOS Ising solvers for real-time, low-energy text summarization on edge devices.
Förderung
- National Science Foundation
- Defense Advanced Research Projects Agency (DARPA) Quantum-Inspired Classical Computing (QuICC) Program through Air Force Research Laboratory
Zitieren
APA
Zeng, Z., Kumar, A., Efe, A., Yin, R., Kim, C. H., Karpuzcu, U. R., & Sapatnekar, S. S. (2026). Extractive Summarization on a CMOS Ising Machine. IEEE Access, 14, 56724-56735. https://doi.org/10.1109/access.2026.3681908
Vancouver
Zeng Z, Kumar A, Efe A, Yin R, Kim CH, Karpuzcu UR, et al. Extractive Summarization on a CMOS Ising Machine. IEEE Access. 2026;14:56724-56735. doi:10.1109/access.2026.3681908
BibTeX
@article{Zeng2026,
author = {Zeng, Ziqing and Kumar, Abhimanyu and Efe, Ahmet and Yin, Ruihong and Kim, Chris H. and Karpuzcu, Ulya R. and Sapatnekar, Sachin S.},
title = {Extractive Summarization on a CMOS Ising Machine},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {56724--56735},
doi = {10.1109/access.2026.3681908}
}
EUVIMED