Kinga Tshering

Dhaka, Bangladesh

Machine Learning Research

I am Kinga Tshering. I work on

I hold a B.Sc. in Computer Science and Engineering from North South University, where I studied as a recipient of His Majesty's Scholarship.

Publications Research CV CV on Google Drive

6 Papers and preprints

2 Accepted at ICML & EMNLP 2026

2 Courses taught as TA

Research interests

Evaluation and reliability of large language models. Efficient and resource-aware deep learning, including network pruning and model compression. Interpretability of learned representations.

The thread running through all three is the distance between a benchmark number and a model you can actually rely on. A model that scores well by answering a different question than the one it was asked has not solved the task; a network that survives pruning tells you something about which parameters mattered in the first place. I like problems where measuring the thing properly is itself the contribution.

Language model evaluation

Separating instruction-following compliance from answer correctness in small language models. One accuracy score can’t tell you whether a model reasoned badly or just ignored the output format.

ICML 2026 EMNLP 2026 Benchmarking

Efficient deep learning

Sparse subnetworks under the Lottery Ticket Hypothesis, applied to hyperspectral imaging. Not just how far a network compresses, but which spectral bands survive the pruning.

Pruning Model compression Hyperspectral

Decoding and inference

Extending Asymptotic Probability Decoding from extrapolation across model scale to extrapolation across context length, so the method no longer needs a whole family of models at different sizes.

University of Arizona Decoding

Applied and medical imaging

Few-shot open-set recognition for skin lesion classification. The dangerous failure isn’t a wrong label — it’s a confident one for a condition the model has never seen.

Medical imaging Open-set Explainable AI

Open to

I am applying to graduate programs in machine learning and am open to research collaborations. The fastest way to reach me is email.

news

Oct 08, 2026	Compliance–Correctness Decomposition for Post-Training Evaluation of Small Language Models was accepted to the NeurIPS 2026 Workshop on Muslims in Machine Learning (MusIML).
Sep 28, 2026	Joined Thimphu TechPark Limited as a Software Developer in Thimphu, Bhutan.
Aug 21, 2026	From Language Specifications to Executable Turing Machines was accepted to Findings of the ACL: EMNLP 2026.
Aug 18, 2026	Started a research internship at the University of Arizona, College of Information Science, advised by Prof. Haw-Shiuan Chang, working on decoding methods for language models.

selected publications

ICML 2026

TimeSpot: Benchmarking Geo-Temporal Understanding in Vision–Language Models in Real-World Settings

Azmine Toushik Wasi, Shahriyar Zaman Ridoy, Koushik Ahamed Tonmoy, Kinga Tshering, S. M. Muhtasimul Hasan, Wahid Faisal, Tasnim Mohiuddin, and Md Rizwan Parvez

In Proceedings of the International Conference on Machine Learning (ICML)Accepted , 2026

Bib

@inproceedings{tshering2026timespot,
  title = {TimeSpot: Benchmarking Geo-Temporal Understanding in Vision--Language Models in Real-World Settings},
  author = {Wasi, Azmine Toushik and Ridoy, Shahriyar Zaman and Tonmoy, Koushik Ahamed and Tshering, Kinga and Hasan, S. M. Muhtasimul and Faisal, Wahid and Mohiuddin, Tasnim and Parvez, Md Rizwan},
  booktitle = {Proceedings of the International Conference on Machine Learning (ICML)},
  year = {2026},
}

EMNLP 2026

From Language Specifications to Executable Turing Machines: Evaluating LLMs as Computational Machine Designers

Shahriyar Zaman Ridoy, S. M. Muhtasimul Hasan, Azmine Toushik Wasi, Kinga Tshering, Koushik Ahamed Tonmoy, and Md Mosaddek Khan

In Findings of the Association for Computational Linguistics: EMNLPAccepted , 2026

Bib

@inproceedings{tshering2026tmeval,
  title = {From Language Specifications to Executable Turing Machines: Evaluating LLMs as Computational Machine Designers},
  author = {Ridoy, Shahriyar Zaman and Hasan, S. M. Muhtasimul and Wasi, Azmine Toushik and Tshering, Kinga and Tonmoy, Koushik Ahamed and Khan, Md Mosaddek},
  booktitle = {Findings of the Association for Computational Linguistics: EMNLP},
  year = {2026},
}

Zorig Melong

The Screen-Time Trap: An Explainable Machine Learning Framework for Pediatric Screen-Time Overuse Prediction

Kinga Tshering, Mushfika Hossain Piya, Ashraf Sharif, and Marzan Islam

Zorig MelongPublished , 2026

Bib

@article{tshering2026screentime,
  title = {The Screen-Time Trap: An Explainable Machine Learning Framework for Pediatric Screen-Time Overuse Prediction},
  author = {Tshering, Kinga and Hossain Piya, Mushfika and Sharif, Ashraf and Islam, Marzan},
  journal = {Zorig Melong},
  volume = {9},
  number = {1},
  year = {2026},
}