Yawei Li

Democratizing AI for pervasive applications.

prof_pic.jpg

50 Nanyang Avenue

Singapore 639798

Open positions: PhD, postdoc, and NTU student projects. Join us →

I am a Nanyang Assistant Professor at School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore. Previously, I was a Lecturer at ETH Zürich, where I worked with Prof. Luca Benini. I also collaborated with Prof. Radu Timofte, Dr. Michele Magno, and Prof. Ming-Hsuan Yang. I got my PhD degree from the Computer Vision Lab supervised by Prof. Luc Van Gool. During my PhD, I was mentored by Babak Ehteshami Bejnordi, Tijmen Blankevoort, and Amir Habibian at Qualcomm and Rakesh Ranjan at Meta. After my graduation, I worked as a Postdoc at CVL, and part-time at Meta.

My research focuses on AI efficiency, aiming to accelerate computation, reduce memory footprint, and mitigate power consumption in AI systems. This is a fascinating area that drives AI applications across both cloud and edge environments. Addressing this challenge requires delving into the system stack to understand its fundamental units, from input modalities, models, computation mechanisms, mappings to hardware specifications. Through joint optimization across techniques and layers, it becomes possible to achieve extreme computational efficiency.

My research spans the AI system stack. The common thread is that efficiency is rarely won at a single layer, the largest gains come from evolving the algorithm and the hardware that runs it together.

  • Computationally efficient model design: what is the right computational mechanism for a given architecture? I have studied this for CNNs, graph neural networks, vision Transformers (LocalViT, GRL, SemanIR, VRT), and state-space models (MambaIR, FEMBA).
  • Model compression and deployment: how much of a trained model is actually needed? My work spans weight pruning (group sparsity, random pruning, DHP), token pruning (FastVAR), quantization (SliM-LLM, OBR), tensor decomposition (learning filter basis), and low-rank adaptation (IntLoRA), applied to LLMs and diffusion models.
  • Software–hardware co-design: once the model is fixed, the bottleneck moves to how it is mapped onto silicon. I work on serving modern models efficiently both on large clusters (FlatAttention) and on severely constrained edge devices (smart glasses, Retina, TinyTracker).
  • Foundation models for biosignals: biosignals record physiological activity and underpin healthcare, disease monitoring, BCI, HCI and AR/VR. They demand accurate modelling under a tight on-device compute budget — here efficiency is not an optimization but a precondition. This is my most active recent direction (PhysioWave, LUNA, WaveFormer, FEMBA, CEReBrO).

I am recruiting PhD students (Spring and Fall 2027 intakes), research fellows, research associates, and NTU project students (FYP, URECA, MSc by Dissertation). The lab works on AI efficiency across the full stack: embodied AI systems, IC design and EDA, efficient algorithms, and biosignal foundation models. See Join Us for details and how to apply.

News

Nov 10, 2026 PhysioWave-T is released on arxiv! PhysioWave-T is a token-efficient wavelet foundation model for biosignals. Congrats Yanlong!
Oct 15, 2026 I will serve as an Area Chair for ICML 2027.
Sep 28, 2026 PrismQuant is released on arxiv! In this work, we derive optimal null-space rotations for grouped quantizers. Congrats Yanlong!
Sep 25, 2026 One work on floating-point microformat quantization and pruning for efficient MU-MIMO neural receivers is released on arxiv.
Sep 21, 2026 Our paper “TinyMyo: A Tiny Foundation Model for Flexible EMG Signal Processing at the Edge” is accepted to IEEE Transactions on Biomedical Engineering (TBME)! Congrats Matteo!
Sep 21, 2026 Our paper “SPEAR: A System for Post-Quantization Error-Adaptive Recovery Enabling Efficient Low-Bit LLM Serving” is accepted to EuroSys 2027!
Aug 15, 2026 I am invited to serve as an Area Chair for ICLR 2027.
Jul 08, 2026 Our paper “Scalable Attention for 5G NR Channel Estimation” is accepted to PIMRC 2026. Congrats Mahdi! In this paper, we propose Channel Estimation Attention (CHEA), a low-complexity channel estimator for 5G New Radio (5G NR) multi-user multiple-input multiple-output (MU-MIMO).
Jul 01, 2026 I joined the School of Electrical and Electronic Engineering at Nanyang Technological University (NTU), Singapore, as a Nanyang Assistant Professor! I am recruiting PhD students, research fellows, research associates, and NTU project students. Check the open positions here.
Jun 13, 2026 Our paper “S-CEReBrO: Breaking the Memory Barrier in Continuous EEG Monitoring” is accepted to MICCAI 2026. Congrats Glenn! Hard work paid off. Another paper “PanLUNA: An Efficient and Robust Query-Unified Multimodal Model for Edge Biosignal Intelligence” is accepted to AICAS 2026. Congrats Marija! Both papers demonstrate our strength in developing efficient biosignal foundation models for edge intelligence.
May 16, 2026 Two papers are released on arxiv. RDKV proposes a rate-distortion bit allocation method for KV cache compression. In this paper, we differentiate the compression-tolerant components from the sensitive components and propose a joint token eviction and quantization method. In Beyond GSD-as-Token, we propose a parameter-efficient fine-tuning framework remote sensing VLMs.
Apr 30, 2026 Two papers are accepted to ICML 2026! The first paper proposes GRACE, a confidence-based distillation method for efficient VLMs. The second paper proposes Brain Signal Rendering, a encoder that renders EEG signals as a Spectrum Video and learns robust, layout agnostic spatial-temporal representations for better generalization across EEG datasets.
Apr 13, 2026 Our paper “FEMBA on the Edge: Physiologically-Aware Pre-Training, Quantization, and Deployment of a Bidirectional Mamba EEG Foundation Model on an Ultra-low Power Microcontroller” is accepted to IEEE Transactions on Biomedical Engineering (TBME)! Congrats Anna!
Apr 07, 2026 One paper that attacks the memory wall for deploying visual autoregressive models on edge GPUs is accepted to ICPR 2026. Please check the paper here.
Feb 03, 2026 Two papers are on LLM and VLM compression are released on arxiv! One paper on robotic grasping and placement with EEG is accepted to ICRA 2026! Check the GRACE paper here and the VQRound paper here.
Feb 02, 2026 Two papers are accepted to ICLR 2026! The first paper is on efficient degradation-agnostic image restoration. The second paper is on joint LLM quantization and pruning.
Jan 20, 2026 I’m awarded the prestigious Nanyang Assistant Professorship. I will be joining Nanyang Technological University soon. Stay tuned for upcoming opportunities for Research Assistant, Research Associate, and Research Fellow positions.
Sep 18, 2025 Three papers on foundation models are accepted to NeurIPS 2025! One paper is on multimodal biosignal foundation model, one is on topology-agnostic EEG foundation model, and another one on segment anything in camouflaged videos with SAM2. Congrats Yanlong, Berkay, and Yuli.
Sep 14, 2025 One work on joint quantization and sparsification is released on arxiv. In this work, we propose an error compensation method to reconcile the conflicting requirements of quantization and pruning. The proposed method delivers up to 4.72x speedup and 6.4x memory reduction compared to the FP16-dense baseline.
Aug 05, 2025 One work Waveformer is accepted to NER 2025! In this paper, Wavelet transform is introduced to develop a transformer model for gesture recognition with EMG signals.
Aug 02, 2025 One paper on AI/ML acceleration on edge device is accepted to CASES in conjunction with ESWEEK!
Aug 01, 2025 One paper is accepted to GLOBECOM 2025! The paper introduces a compute and memory efficient model for 5G receiver on the edge devices.
Jun 25, 2025 One paper on the acceleration of visual autoregressive models is accepted to ICCV 2025! Check the paper here.
May 11, 2025 One paper on dataflow and accelerator co-design is accepted to ISVLSI 2025! Check the paper here.
May 01, 2025 Two papers on model quantization are accepted to ICML 2025! IntLoRA proposes an integral low-rank adaption method for quantized diffusion models. After low-rank adaptation, all the weights are converted to integers. SliM-LLM proposes a mixed precision quantization method for LLMs.
Apr 29, 2025 Two papers on biosignal foundation models are accepted to EMBC 2025! One paper introduces FEMBA, a state-space model for efficient and scalable EEG analysis. The other paper develops a model for blood pressure estimation with ECG and PPG signals.
Feb 26, 2025 Three papers are accepted to CVPR 2025! One paper is on state space models, one is on foundation models for Markush structures, and another one on semantic and sequential alignment for vision-language models.
Jan 18, 2025 We are organizing the 10th NTIRE workshop and challenge on efficient image super-resolution and image denoising.
Sep 26, 2024 One paper on Graph Attention in Transformers is accepted to NeurIPS 2024! Check the paper here.
Sep 25, 2024 One paper on masked autoencoder for point clouds is accepted to ACCV 2024! Check the paper here.
Aug 30, 2024 TinyTracker is accepted to IEEE Sensors. Check the paper here.
Jul 01, 2024 One paper on Open-Vocabulary Video Segmentation is accepted to ECCV 2024! Check the paper here.
Apr 30, 2024 LSDIR dataset is accepted as a workshop paper at CVPR 2023. Check the paper here.
Mar 12, 2024 One paper is accepted to ICME. Check the paper here.
Mar 01, 2024 We open PhD positions on Secure Machine Learning on RISC-V Servers and Accelerators at ETH Zürich. More info here.

Selected Publications

  1. Arxiv
    2026_arxiv_prismquant.png
    PrismQuant: Optimal Null-Space Rotations for Grouped Quantizers
    Yanlong Chen, Yining Chen, Song Zhang, Amirhossein Habibian, and Yawei Li
    arXiv preprint arXiv:2609.32429, 2026
  2. ICML
    2026_arxiv_grace.png
    Gated Relational Alignment via Confidence-based Distillation for Efficient VLMs
    Yanlong Chen, Amirhossein Habibian, Luca Benini, and Yawei Li
    In Proceedings of International Conference on Machine Learning (ICML), 2026
  3. ICLR
    2026_iclr_obr.png
    Optimal Brain Restoration for Joint Quantization and Sparsification of LLMs
    Hang Guo, Luca Benini, and Yawei Li
    In The Fourteenth International Conference on Learning Representations (ICLR), 2026
  4. NeurIPS
    2025_neurips_physiowave.jpg
    PhysioWave: A Multi-Scale Wavelet-Transformer for Physiological Signal Representation
    Yanlong Chen, Mattia Orlandi, Pierangelo Maria Rapa, Simone Benatti, Luca Benini, and Yawei Li
    In Advances in Neural Information Processing Systems (NeurIPS), 2025
  5. NeurIPS
    2025_neurips_luna.jpg
    LUNA: Efficient and Topology-Agnostic Foundation Model for EEG Signal Analysis
    Berkay Döner, Thorir Mar Ingolfsson, Luca Benini, and Yawei Li
    In Advances in Neural Information Processing Systems (NeurIPS), 2025
  6. ICCV
    2025_iccv_fastvar.jpg
    FastVAR: Linear Visual Autoregressive Modeling via Cached Token Pruning
    Hang Guo, Yawei Li, Taolin Zhang, Jiangshan Wang, Tao Dai, Shu-Tao Xia, and Luca Benini
    In Proceedings of the IEEE/CVF International Conference on Computer Vision, 2025
  7. ISVLSI
    2025_isvlsi_flatattention.jpg
    FlatAttention: Dataflow and Fabric Collectives Co-Optimization for Efficient Multi-Head Attention on Tile-Based Many-PE Accelerators
    Chi Zhang, Luca Colagrande, Renzo Andri, Thomas Benz, Gamze Islamoglu, Alessandro Nadalini, Francesco Conti, Yawei Li, and Luca Benini
    In IEEE Computer Society Annual Symposium on VLSI, 2025
  8. CVPR
    2025_cvpr_mambairv2.jpg
    MambaIRv2: Attentive State Space Restoration
    Hang Guo, Yong Guo, Yaohua Zha, Yulun Zhang, Wenbo Li, Tao Dai, Shu-Tao Xia, and Yawei Li
    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2025
  9. NeurIPS
    2024_neurips_semanir.jpg
    Sharing Key Semantics in Transformer Makes Efficient Image Restoration
    Bin Ren, Yawei Li, Jingyun Liang, Rakesh Ranjan, Mengyuan Liu, Rita Cucchiara, Luc Van Gool, Ming-Hsuan Yang, and Nicu Sebe
    In Advances in Neural Information Processing Systems, 2024
  10. ECCV
    2024_eccv_smart_glasses.jpg
    Ultra-Efficient On-Device Object Detection on AI-Integrated Smart Glasses with TinyissimoYOLO
    Julian Moosmann, Pietro Bonazzi, Yawei Li, Sizhen Bian, Philipp Mayer, Luca Benini, and Michele Magno
    In Proceedings of the European conference on computer vision Workshops, 2024
  11. CVPR
    2024_cvpr_retina.png
    Retina: Low-Power Eye Tracking with Event Camera and Spiking Hardware
    Pietro Bonazzi, Sizhen Bian, Giovanni Lippolis, Yawei Li, Sadique Sheik, and Michele Magno
    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2024
  12. IROS
    2023_iros_localvit.png
    LocalViT: Analyzing Locality in Vision Transformers
    Yawei Li, Kai Zhang, Jiezhang Cao, Radu Timofte, Michele Magno, Luca Benini, and Luc Van Gool
    In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, 2023
  13. IEEE Sensors
    2023_sensors_tinytracker.png
    TinyTracker: Ultra-Fast and Ultra-Low-Power Edge Vision for In-Sensor Gaze Estimation
    Pietro Bonazzi, Thomas Ruegg, Sizhen Bian, Yawei Li, and Michele Magno
    In Proceedings of IEEE Sensors, 2023
  14. CVPR
    2023_cvpr_grl.jpg
    Efficient and Explicit Modelling of Image Hierarchies for Image Restoration
    Yawei Li, Yuchen Fan, Xiaoyu Xiang, Denis Demandolx, Rakesh Ranjan, Radu Timofte, and Luc Van Gool
    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023
  15. CVPR
    2022_cvpr_random_pruning.png
    Revisiting Random Channel Pruning for Neural Network Compression
    Yawei Li, Kamil Adamczewski, Wen Li, Shuhang Gu, Radu Timofte, and Luc Van Gool
    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022
  16. ECCV
    2020_eccv_dhp.png
    DHP: Differentiable Meta Pruning via HyperNetworks
    Yawei Li, Shuhang Gu, Kai Zhang, Luc Van Gool, and Radu Timofte
    In Proceedings of the European Conference on Computer Vision, 2020
  17. CVPR
    2020_cvpr_group_sparsity.png
    Group Sparsity: The Hinge Between Filter Pruning and Decomposition for Network Compression
    Yawei Li, Shuhang Gu, Christoph Mayer, Luc Van Gool, and Radu Timofte
    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020