Hi! I’m Shubhra, an incoming PhD candidate at KTH University in Stockholm, Sweden, advised by Dr. David Broman. I’ll be graduating in June with my B.S. + M.S. from Stanford studying Computer Science (AI track). My main research focus is improving math and code reasoning skills in AI models using neurosymbolic techniques. I do research at Stanford’s Computation and Cognition Lab, advised by Dr. Noah Goodman. I believe that understanding human thinking is important to exploring how we can design better reasoning models. To do so, I’m working with the students from the CoCoSci group at MIT to understand what makes specific math problems interesting to humans. I’m also interested in education and exploring how LLMs can help math teachers better scaffold their lessons under Dr. Dora Demszky at the EduNLP Lab.

Publications

2024

MathCAMPS: Fine-grained Synthesis of Mathematical Problems From Human Curricula

NeurIPS Math-AI Workshop 2024
Shubhra Mishra*, Gabriel Poesia*, Belinda Mo, Noah Goodman
[Website] [arXiv]

An Evaluation Benchmark for Autoformalization

ICLR 2024
Aryan Gulati*, Devanshu Ladsaria*, Shubhra Mishra*, Jasdeep Sidhu*, Brando Miranda
[arXiv]

Families of Harris Graphs

In submission
Shubhra Mishra, Doug Shaw, Francesca Gandini
[Website] [arXiv]

Projects

2024

🏆 Can Symbolic Scaffolding and DPO Enhance Mathematical Problem-Solving Skills in LLMs?

CS 329H: Machine Learning from Human Preferences
🏆 Outstanding Project Award for CS 329H
Paper coming soon!

Improving Counting Abilities in Stable Diffusion Models

CS 468: Topics in Geometric Computing - 3D and 4D Foundation Models
[Paper]

🏆 Self-Improvement for Math Problem-Solving in Small Language Models

CS 224N: Natural Language Processing with Deep Learning
🏆 Outstanding Project Award for CS 224N
[Paper]

GAMMAS: Improving Mathematical Reasoning in Vision Language Models Through Synthetic Data Generation

CS 231N: Deep Learning for Computer Vision
[Paper]