Sudarshan Regmi

I am a second-year PhD student at Dartmouth College. I am fortunate to be advised by Professor Andrew Campbell. My research interests revolve around trustworthy machine learning. I obtained my Bachelor degree in 2022 from Department of Electronics and Computer Engineering at Pulchowk Campus, Tribhuvan University.

Email / Scholar / Twitter / Github / Linkedin

True wisdom is recognizing what you know,
and
acknowledging what you don't.

Research

My current research work centers on LLMs / generative modeling for time-series modality.

MontageFlow: Physically Valid High-Fidelity EEG Generation
Sudarshan Regmi et al.
arXiv preprint 2610.14507, 2026
PDF | Code

TimeThink: Eliciting Compositional Reasoning in Timeseries Large Language Models
Sudarshan Regmi et al.
arXiv preprint 2609.13457, 2026
PDF | Code

AdaSCALE: Adaptive Scaling for OOD Detection
Sudarshan Regmi
International Conference on Machine Learning (ICML), 2026
PDF | Code

Image-based Outlier Synthesis With Training Data
Sudarshan Regmi
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026
PDF | Code

ReweightOOD: Loss Reweighting for Distance-based OOD Detection
Sudarshan Regmi et al.
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2024
PDF | Code

T2FNorm: Train-time Feature Normalization for OOD Detection in Image Classification
Sudarshan Regmi et al.
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2024
PDF | Code

For complete list of publications, visit Google Scholar

Academic Services

Conference reviewer: CVPR, MICCAI, ICCV, NeurIPS

Journal reviewer: Pattern Recognition, Medical Image Analysis

Miscellaneous

Outside of research, I like watching/playing cricket and hiking. I also frequently enjoy revisiting Tolkien's Middle-earth. I enjoy comedy shows. After all, what is life without a touch of humor? 😉

Lucky welcomes you!