Abhishek Panigrahi

I am a fifth year graduate student in the Computer Science department at Princeton University, fortunate to be advised by Prof. Sanjeev Arora. Previously, I was a Research Fellow at Microsoft Research Lab - India where I worked with Dr. Harsha Vardhan Simhadri and Dr. Navin Goyal. Prior to the fellowship, I attended IIT Kharagpur where I obtained my B.Tech. in Computer Science and Engineering in 2018.

I am an Apple AI/ML Ph.D. scholar and a Siebel scholar for the academic year 2025-26.

Email / CV / Scholar /

I will be on the job market for 2026. Please reach out if you think my background and experience could be a good fit for your organization.

Research

My research aims to move beyond the paradigm of “more compute, better results” toward “targeted compute, stronger generalization.” I study what makes language models truly adaptable to more difficult tasks beyond human supervision -- which I refer to as strong generalization. My goal is to identify and optimize the components of the training pipeline that give rise to robust, adaptable reasoning in language models.

Representative Works

In Good GRACES: Principled Teacher Selection for Knowledge Distillation
Abhishek Panigrahi, Bingbin Liu, Sadhika Malladi, Sham Kakade, Surbhi Goel
In submission

STAT: Skill-Targeted Adaptive Training
Yinghui He*, Abhishek Panigrahi*, Yong Lin, Sanjeev Arora
In submission
arXiv / code

On the Power of Context-Enhanced Learning in LLMs
Xingyu Zhu*, Abhishek Panigrahi*, Sanjeev Arora
International Conference on Machine Learning (ICML 2025) (Spotlight)
arXiv / code

Generalizing from SIMPLE to HARD Visual Reasoning: Can We Mitigate Modality Imbalance in VLMs?
Simon Park*, Abhishek Panigrahi*, Catherine Cheng*, Dingli Yu, Anirudh Goyal, Sanjeev Arora
International Conference on Machine Learning (ICML 2025)
arXiv / code

Progressive distillation induces an implicit curriculum
Abhishek Panigrahi*, Bingbin Liu*, Sadhika Malladi, Andrej Risteski, Surbhi Goel
International Conference on Learning Representations (ICLR 2025) (Oral) Openreview / code / blog

Efficient stagewise pretraining via progressive subnetworks
Abhishek Panigrahi*, Nikunj Saunshi*, Kaifeng Lyu, Sobhan Miryoosefi, Sashank Reddi, Satyen Kale, Sanjiv Kumar
International Conference on Learning Representations (ICLR 2025)
Openreview

Task-specific skill localization in fine-tuned language models
Abhishek Panigrahi*, Nikunj Saunshi*, Haoyu Zhao, Sanjeev Arora
International Conference of Machine Learning (ICML 2023)
PMLR / code

Understanding Gradient Descent on the Edge of Stability in Deep Learning
Sanjeev Arora, Zhiyuan Li, Abhishek Panigrahi (alphabetical)
International Conference of Machine Learning (2022) (Spotlight)
PMLR / code

Other representative works

Trainable Transformer in Transformer
Abhishek Panigrahi*, Sadhika Malladi*, Mengzhou Xia, Sanjeev Arora
International Conference of Machine Learning (ICML 2024)
Openreview / Code

Do transformers parse while predicting the masked word?
Haoyu Zhao*, Abhishek Panigrahi*, Rong Ge, Sanjeev Arora
Empirical Methods in Natural Language Processing (EMNLP 2023)
Arxiv

Effect of Activation Functions on the Training of Overparametrized Neural Nets
Abhishek Panigrahi*, Abhishek Shetty*, Navin Goyal
International Conference on Learning Representations (ICLR 2020)
Openreview

Word2Sense: Sparse interpretable word embeddings
Abhishek Panigrahi, Harsha Vardhan Simhadri, Chiranjib Bhattacharyya
Association for Computational Linguistics (ACL 2019) (Oral)
ACL / code