Kaustubh Ponkshe

PhD student, EPFL

Joined

October 2023

Names

Kaustubh Ponkshe (Preferred)

Emails

****@gmail.com (Confirmed)

Personal Links

Homepage

Google Scholar

Career & Education History

PhD student

EPFL (epfl.ch)

2025 – 2029

Undergrad student

Indian Institute of Technology Bombay, Indian Institute of Technology, Bombay (cse.iitb.ac.in)

2019 – 2024

Advisors, Relations & Conflicts

PhD Advisor

Martin Jaggi

Present

Expertise

No areas of expertise listed

Publications

[Short] Studying Memorization Dynamics in Large Language Models Across Pre-Training
Kaustubh Ponkshe, Raghav Singhal, Daniele Affinita, Martin Jaggi
- ICLR 2026 Workshop DATA-FM
- Readers: Everyone
Fed-SB: A Silver Bullet for Extreme Communication Efficiency and Performance in (Private) Federated LoRA Fine-Tuning
Raghav Singhal, Kaustubh Ponkshe, Rohit Vartak, Lav R. Varshney, Praneeth Vepakomma
- Accepted by TMLR
- Readers: Everyone
Fed-SB: A Silver Bullet for Extreme Communication Efficiency and Performance in (Private) Federated LoRA Fine-Tuning
Raghav Singhal, Kaustubh Ponkshe, Rohit Vartak, Lav R. Varshney, Praneeth Vepakomma
- ICLR 2026 Conference Withdrawn Submission
- Readers: Everyone
Initialization using Update Approximation is a Silver Bullet for Extremely Efficient Low-Rank Fine-Tuning
Kaustubh Ponkshe, Raghav Singhal, Eduard Gorbunov, Alexey Tumanov, Samuel Horváth, Praneeth Vepakomma
- ICLR 2026 Conference Withdrawn Submission
- Readers: Everyone
ABBA-Adapters: Efficient and Expressive Fine-Tuning of Foundation Models
Raghav Singhal, Kaustubh Ponkshe, Rohit Vartak, Praneeth Vepakomma
- ICLR 2026 Poster
- Readers: Everyone
Safety Subspaces are Not Linearly Distinct: A Fine-Tuning Case Study
Kaustubh Ponkshe, Shaan Shah, Raghav Singhal, Praneeth Vepakomma
- ICLR 2026 Poster
- Readers: Everyone
Safety Subspaces are Not Distinct: A Fine-Tuning Case Study
Kaustubh Ponkshe, Shaan Shah, Raghav Singhal, Praneeth Vepakomma
- NeurIPS Lock-LLM Workshop 2025 Poster
- Readers: Everyone
TokenSwap: A Lightweight Method to Disrupt Memorized Sequences in LLMs
Kaustubh Ponkshe, Parjanya Prajakta Prashant, Babak Salimi
- NeurIPS 2025 Workshop GenProCC
- Readers: Everyone
Safety Subspaces are Not Distinct: A Fine-Tuning Case Study
Shaan Shah, Kaustubh Ponkshe, Raghav Singhal, Praneeth Vepakomma
- INTERPLAY
- Readers: Everyone
Fed-SB: A Silver Bullet for Extreme Communication Efficiency and Performance in (Private) Federated LoRA Fine-Tuning
Raghav Singhal, Kaustubh Ponkshe, Rohit Vartak, Lav R. Varshney, Praneeth Vepakomma
- ES-FoMo III
- Readers: Everyone

View all 16 publications