Prasoon Bajpai

I am a pre-doctoral researcher at Google DeepMind, where I am advised by Shachi Dave and Partha Talukdar. I currently work on recursive self-improvement. I have also worked on cross-lingual transfer and pre-training evaluations, as well as on open-endedness — applying it to automated red-teaming. Before DeepMind, I completed my bachelor's and master's in Mathematics and Computing at IIT Delhi, where my master's thesis was supervised by Dr. Tanmoy Chakraborty.

Outside of research, my big interest is strength training, and I am currently focused on improving my bouldering. I am also drawn to endurance sports, particularly long-distance running. Beyond that, I watch a lot of English Premier League football — I support Chelsea — and I am an ardent Fantasy Premier League player.

Scholar  /  Twitter

profile photo

Thoughts

  • The first one is coming soon.

Research

I am interested in multilingual NLP, long-context reasoning, test-time scaling, interpretability, and the evaluation of large language models.

Are Multilingual Models Actually Improving? Isolating True Cross-Lingual Transfer
Prasoon Bajpai, Eleftheria Briakou, Colin Cherry, Preethi Jyothi, Vihari Piratla
EMNLP, 2026
arXiv
Interpreting and Controlling Model Behavior via Constitutions for Atomic Concept Edits
Neha Kalibhat, Zi Wang, Prasoon Bajpai, Drew Proud, Wenjun Zeng, Been Kim, Mani Malek
AISTATS, 2026
arXiv
Can LLMs Reason over Extended Multilingual Contexts? Towards Long-Context Evaluation Beyond Retrieval and Haystacks
Amey Hengle, Prasoon Bajpai, Soham Dan, Tanmoy Chakraborty
EACL, 2026
arXiv
Multilingual Test-Time Scaling via Initial Thought Transfer
Prasoon Bajpai, Tanmoy Chakraborty
arXiv, 2025
arXiv
Information Anxiety in Large Language Models
Prasoon Bajpai, Sarah Masud, Tanmoy Chakraborty
arXiv, 2024
arXiv
Multilingual Needle in a Haystack: Investigating Long-Context Behavior of Multilingual Large Language Models
Amey Hengle, Prasoon Bajpai, Soham Dan, Tanmoy Chakraborty
NAACL, 2025
arXiv
Can LLMs Replace Neil deGrasse Tyson? Evaluating the Reliability of LLMs as Science Communicators
Prasoon Bajpai, Niladri Chatterjee, Subhabrata Dutta, Tanmoy Chakraborty
EMNLP, 2024
arXiv

Website template adapted from Jon Barron.