Yash Mali

👋 I am an undergraduate student at the University of British Columbia. I am broadly interested in understanding what models learn and studying their robustness under distribution shift. I am also interested in the applications of AI in science and society.

Email: ymali@student.ubc.ca / Google Scholar / LinkedIn / GitHub / CV / arXiv

Research

LookSharp: Attention Entropy Minimization for Test-Time Adaptation: ICLR Workshops TTU & CAO 2026

Yash Mali, Evan Shelhamer

We minimize the entropy of CLS-to-patch attention in the final layer as a novel test-time loss, encouraging the model to maintain focused attention on shifted data. Paper / Code / Poster

prAxIs UBC: Developing AI education modules in Faculty of Arts Courses: Conference on Teaching and Research in Economic Education Workshop, 2026

Jonathan Graves, Laura Nelson, and prAxIs Contributors (including Yash Mali)

AI education in the Faculty of Arts. Building modules that teach students how to use AI to solve real problems in the arts. Website

A Chatbot for the Management of Bipolar Disorder

Yash Mali, Zejiao Zeng, Kayoung Heo, Grace Zhang, Jincheng Chen, Kamyar Keramatian, Gayatri Saraf, Marco Solmi, Edwin Tam, Sagar V. Parikh, Ayal Schaffer, Serge Beaulieu, Raymond Ng, Lakshmi N. Yatham, John-Jose Nunez

Agentic LLM system for doctors that provides fast and accurate medical recommendations with reference from reliable sources (Co-op, May 25 - Aug 25). Paper / Summary

Investigating fine-tuning versus zero-shot learning for general large language models when predicting cancer survival from initial oncology consultation documents: ESMO Digital Oncology, 2026

Tejas Phaterpekar, Zejiao Zeng, Yash Mali, Bonnie Leung, Carol Ho, Raymond Ng, Alan T Bates, John-Jose Nunez

LLM fine-tuning for survival prediction. (Co-op, May 25 - Aug 25). Paper

A particle cohort study (ParCS) of the impact of glucose and sucrose solutions on the kinetics of starch gelatinization: Food Hydrocolloids, 2026

Lily Santos O’ Keefe, Yash Mali, John Frostad

Contributed particle tracking software for tracking starch particles to study the kinetics of starch gelatinization (Co-op, May 24 - Aug 24). Paper / Code

Other Research

Undergraduate Thesis

Yash Mali, Tina Behnia, Christos Thrampoulidis

Investigated how noise and diversity in the training data affect the internal representations and factual recall of autoregressive language models.

Other Highlights