Vedant Palit

I am a fourth year undergraduate student at IIT Kharagpur pursuing a BTech in Industrial & Systems Engineering and an MTech in Financial Engineering.

I take interest in machine and deep learning, particularly in the domain of natural language processing and computer vision. Recently I have been delving into the specifics of mechanistic interpretability. I also enjoy quizzing, reading sci-fi books and watching anime.

Feel free to drop me an e-mail if you want to chat with me!

Email  /  CV  /  Scholar  /  Blog  /  Github

profile photo

📢 News & Updates
  • [Jan '25] Our Paper "What Do VLMs NOTICE? A Mechanistic Interpretability Pipeline for Noise-free Text-Image Corruption and Evaluation" has been accepted to NAACL 2025.
  • [Dec '24] Captained the team of IIT Kharagpur to Gold at Adobe Image Classification Challenge (Inter IIT Tech Meet 2024).
  • [Oct '24] Served as a Reviewer at ICLR 2024.
  • [Sept '24] Our paper "WellDunn: On the Robustness and Explainability of Language Models and Large Language Models in Identifying Wellness Dimensions" has been accepted at the EMNLP BlackBox NLP Workshop 2024.
  • [Jun '24]Served as a Reviewer at the ACM SIGKDD Workshop 2024 .
  • [Feb '24] Started as an Undergraduate Researcher at the Eickhoff AI Lab, Brown University, USA.
  • [Dec '23] Started as a Research Intern at Inria Centre, University of Lille, France.
  • [July '23] Our paper "Towards Vision-Language Mechanistic Interpretability: A Causal Tracing Tool for BLIP" has been accepted at the ICCV CLVL Workshop 2023.
  • [June '23] Our paper "Knowledge Graph Guided Semantic Evaluation of Language Models For User Trust" was accepted at IEEE CAI 2023.
  • [Nov '22] Started as a Research Intern at University of Maryland, Baltimore.
  • [Dec '21] Accepted into IIT Kharagpur.

🔬 Research
Forgotten Polygons: Multimodal Large Language Models are Shape-Blind
William Rudman*, Michal Golovanesky*, Amir Bar, Vedant Palit, Yann LeCun, Carsten Eickhoff, Ritambhara Singh
Under Review at ACL 2025
Repo / Paper

What Do VLMs NOTICE? A Mechanistic Interpretability Pipeline for Noise-free Text-Image Corruption and Evaluation
Michal Golovanesky*, William Rudman*, Vedant Palit, Ritambhara Singh, Carsten Eickhoff
NAACL 2025
Repo / Paper

WellDunn: On the Robustness and Explainability of Language Models and Large Language Models in Identifying Wellness Dimensions
Ali Mohammadi, Edward Raff, Jinendra Malekar, Vedant Palit, Francis Ferraro, Manas Gaur
EMNLP 2024 (7th BlackBox NLP Workshop)
Repo / Paper

Towards Vision-Language Mechanistic Interpretability: A Causal Tracing Tool for BLIP
Vedant Palit*, Rohan Pandey*, Aryaman Arora, Paul Pu Liang
ICCV 2023 (5th CLVL Workshop)
Repo / Paper

Knowledge Graph Guided Semantic Evaluation of Language Models For User Trust
Kaushik Roy, Tarun Garg, Vedant Palit, Yuxin Zi, Vignesh Narayanan, Amit Sheth
IEEE Conference on Artificial Intelligence 2023
Repo / Paper

Exploring The Potential of Large Language Models for Assisting with Mental Health Diagnostic Assessments
Kaushik Roy, Harshul Surana, Darssan Eswaramoorthi, Yuxin Zi, Vedant Palit, Ritvik Garimella, Amit Sheth
Under Review in the ACM Health Journal
Repo / Preprint

🛠️ Projects
Knowledge Graph Generation from Scraped Text
AIISC Project
Repository

Created an automated text extraction system, for custom information extraction from an input wikipedia page followed by entity generation through parsing, named entity correlation and KG generation.

Image Stitching and RANSAC Line Fitting
Task for Autonomous Ground Vehicles Group, IIT KGP
Repository

Constructed a pipeline of the image stitching through brute-force matching as well as KNN Association of keypoints in two images of the same scene taken from close yet different angles.

Kalman Filter Trajectory Determination
Task for Autonomous Ground Vehicles Group, IIT KGP
Repository

Developed a trajectory determination program based on the kalman filter algorithm utilising information from acceleration and position sensors to predict the next position of a vehicle.

✍️ Blogs