Dong Wang (王栋)

I am a PhD student supervised by Olga Saukh at Embedded Learning and Sensing Systems (ELSS) Group, Institute for Technical Informatics, TU Graz, Austria. My research interests include efficient machine learning, on-device and embedded machine learning, deep learning, computer vision, efficient pre-training, and the Internet of Things (IoT).

News

• [ Jan. 2026] 🚀 Our paper Cut Less, Fold More: Model Compression through the Lens of Projection Geometry was accepted at ICLR'26.

• [ Jan. 2026] 🚀 Our paper FOSL: A Foldable Sparse-and-Low-Rank Method for Efficient LLM Pre-training was accepted at CPAL'2026 (Recent Spotlight Track).

• [ Jan. 2026] 🚀 Our paper GRAIL: Post-hoc Compensation by Linear Reconstruction for Compressed Networks was accepted at CPAL'2026 (Proceedings Track).

• [ Nov. 2025] Two workshop papers accepted by EurIPS.

• [ Apr. 2025] I presented the model folding paper at ICLR'25 in Singapore.

• [ Jan. 2025] 🚀 Our paper Forget the Data and Fine-Tuning! Just Fold the Network to Compress was accepted at ICLR'25

• [ Aug. 2024] I gave an oral presentation about Subspace-Configurable Networks at CoLLAs'24 in Pisa, Italy🇮🇹. See pictures here📷

• [ Jun. 2024] I will attend DENISE Summer School.

• [ May. 2024] I will present Subspace-Configurable Networks at the PML4LRS workshop at ICLR'24.

• [ Apr. 2024] 🚀 Our paper presenting Subspace-Configurable Networks was accepted (oral) at CoLLAs'24

• [ Feb. 2024] I presented my research at GraML.

• [ Nov. 2022] I started my PhD at TU Graz, Austria.

• [ Jun. 2021] My master's thesis was approved by Uppsala University.

• [ Jan. 2021] I started my master's thesis at Ericsson.

• [ Jun. 2020] SemiFlow: A deep learning framework with automatic differentiation and automatic shape inference 1.0 was released.

Education

Work Experience

Publication

O. Saukh, D. Wang, H. Šikić, Y. Cheng, L. Thiele, Cut Less, Fold More: Model Compression through the Lens of Projection Geometry, International Conference on Learning Representations (ICLR), 2026, accepted.

D. Wang, F. Corti, Y. Cheng, O. Saukh, FOSL: A Foldable Sparse-and-Low-Rank Method for Efficient LLM Pre-training, Conference on Parsimony and Learning (CPAL), Recent Spotlight Track, 2026, accepted.

W. Tang, D. Wang, L. Thiele, O. Saukh, GRAIL: Post-hoc Compensation by Linear Reconstruction for Compressed Networks, Conference on Parsimony and Learning (CPAL), Proceedings Track, 2026, accepted.

D. Wang, H. Šikić, L. Thiele, O. Saukh, Model Folding: A Unified Approach to Post-training Compression and Efficient Pre-training, Workshop on Energy-Efficient AI: Models, Algorithms, and Hardware for Sustainable Intelligence @EurIPS, 2025, accepted.

O. Saukh, D. Wang, H. Šikić, Y. Cheng, L. Thiele, Cut Less, Fold More: Model Compression through the Lens of Projection Geometry, Workshop on Rethinking AI: Efficiency, Frugality, and Sustainability, @EurIPS, 2025, accepted.

D. Wang, H. Šikić, L. Thiele, O. Saukh, Forget the Data and Fine-Tuning! Just Fold the Network to Compress, International Conference on Learning Representations, 2025 (ICLR)

D. Wang, O. Saukh, X. He, L. Thiele, Subspace-Configurable Networks, Third Conference on Lifelong Learning Agents (CoLLAs), 2024. Accepted, to appear. ICLR Workshop on Practical ML for Limited/Low Resource Settings, 2024. To appear.

D. Wang, and W. Ji, NLP-based social IoT service framework: design and implementation, China Computer & Communication, vol.14, pp.74-77, 2018. Published

Thesis

Bachelor Thesis Dong Wang. Research on object detection algorithm based on GAN. Bachelor's thesis, Xi'an University of Science and Technology, Internet of Things, June 2018. Master Thesis Wang, D. (2021). Low-latency object detection on the edge-cloud: AprilTag-assisted object detection and positioning.

Patent

Dong Wang, “Intelligent somatosensory navigation device”, China Patent CN206573128U, issued Oct 20, 2017

YouTube: An introduction to Intelligent Somatosensory Navigation Device: Suzaku

Awards