Scholar
Pratyusha Sharma
Google Scholar ID: RGiCLUgAAAAJ
PhD Student, Massachusetts Institute of Technology
Citations & Impact
Citations
1,561
H-index
16
i10-index
17
Publications
20
Co-authors
71
list available
Publications
Resume (English only)
Academic Achievements
- Publications: 1. LoRA vs Full Fine-tuning: An Illusion of Equivalence (NeurIPS, 2025); 2. Efficient LLM Adaptation Using a Single Gradient Step on 100 Samples (NeurIPS, 2025); 3. Contextual and combinatorial structure in sperm whale vocalisations (Nature Communications, 2024); 4. The Truth is in There: Improving Reasoning in Language Models with Layer-Selective Rank Reduction (ICLR 2024); Preprints: WhaleLM: Finding Structure and Information in Sperm Whale Vocalizations and Behavior with Machine Learning (bioRxiv, 2024).
Research Experience
- Incoming Assistant Professor at NYU in Computer Science and the Center for Data Science; Senior Research Scientist at Microsoft Research; spent some fun summers at Microsoft Research NYC and NVIDIA Seattle Robotics Lab.
Education
- PhD in EECS at MIT, advised by Prof. Antonio Torralba and Prof. Jacob Andreas; one-year research internship at Carnegie Mellon University's Robotics Institute, advised by Prof. Abhinav Gupta; undergraduate from Indian Institute of Technology, Delhi.
Background
- Research interests: the interplay between language (structure more than form), reasoning, and sequential decision making; goal to understand systems that exhibit broadly intelligent behaviors (AI systems and biological organisms) and build better AI systems. Recent research includes: 1. Developing algorithms to understand the structure of solutions artificial neural networks implement and how it affects generalization; 2. Understanding the complexity and structure of naturally arising animal communication systems in the wild; 3. How language and natural-language-like structures can support effective reasoning and planning in embodied agents and robots.
Miscellany
- Recruiting PhD students for a start in Fall 2026; recent news: invited talks at various workshops and conferences, including ICML Workshop on Machine Learning for Audio, Biennial Conference on the Biology of Marine Mammals, etc.; research featured in National Geographic Magazine; participated in discussions and perspective articles, such as in The New York Times and The Daily Podcast on contextual and combinatorial structure in sperm whale vocalisations.
