Profiles2021-2026
Qianxiao Li
Department of Mathematics National University of Singapore Institute for Functional Intelligent Materials National University of Singapore Department of Mathematics National University of Singapore Ce
18Papers
15Last author
19Unique co-authors
0First author
0Solo
First and last author come from byline position, which some venues assign alphabetically.
Recent papers18
- Machine Unlearning under Retain–Forget EntanglementAuthor 3 of 4ICLR 2026 · 11 citations
- The Effect of Attention Head Count on Transformer ApproximationAuthor 5 of 5ICLR 2026 · 5 citations
- Learning Permutation-invariant Macroscopic DynamicsAuthor 3 of 3ICML 2026
- A unified framework for establishing the universal approximation of transformer-type architecturesAuthor 4 of 4NeurIPS 2025 · 2 citations
- Autocorrelation Matters: Understanding the Role of Initialization Schemes for State Space ModelsAuthor 2 of 2ICLR 2025
- BP-Modified Local Loss for Efficient Training of Deep Neural NetworksAuthor 2 of 2ICLR 2025
- Continuity-Preserving Convolutional Autoencoders for Learning Continuous Latent Dynamical Models from ImagesAuthor 3 of 3ICLR 2025
- From Weight-Based to State-Based Fine-Tuning: Further Memory Reduction on LoRA with Parallel ControlAuthor 4 of 4ICML 2025
- Approximation Rate of the Transformer Architecture for Sequence ModelingAuthor 2 of 2NeurIPS 2024 · 32 citations
- StableSSM: Alleviating the Curse of Memory in State-space Models through Stable ReparameterizationAuthor 2 of 2ICML 2024 · 25 citations
- From Generalization Analysis to Optimization Designs for State Space ModelsAuthor 2 of 2ICML 2024 · 12 citations
- Inverse Approximation Theory for Nonlinear Recurrent Neural NetworksAuthor 3 of 3ICLR 2024 · 10 citations
- Parameter-Efficient Fine-Tuning with ControlsAuthor 4 of 4ICML 2024 · 7 citations
- Learning Macroscopic Dynamics from Partial Microscopic ObservationsAuthor 2 of 2NeurIPS 2024 · 4 citations
- On the Curse of Memory in Recurrent Neural Networks: Approximation and Optimization AnalysisAuthor 3 of 3ICLR 2021 · 40 citations
- Amata: An Annealing Mechanism for Adversarial Training AccelerationAuthor 2 of 4AAAI 2021 · 18 citations
- Approximation Theory of Convolutional Architectures for Time Series ModellingAuthor 3 of 3ICML 2021 · 14 citations
- Towards Robust Neural Networks via Close-loop ControlAuthor 2 of 3ICLR 2021 · 5 citations
