My research interests broadly encompass training, alignment, and benchmarking of language models. With the growing integration of AI into our day to day lives, I am passionate about developing AI systems that are safe, robust and reliable.

I am currently part of the CoreLLM research group at ServiceNow where my work focuses on enhancing the overall capabilities of language models. Prior to that, I was an AI Resident at Facebook AI Research, where I worked with Marcus Rohrbach on developing models with Multimodal understanding of vision and language content. During my Masters, I worked on evaluating the robustness and generalization of language models as part of my thesis at IIIT Hyderabad.
More details are available in my cover letter.

Research

Curry-DPO: Enhancing Alignment using Curriculum Learning & Ranked Preferences
Pulkit Pattnaik, Rishabh Maheshwary, Kelechi Ogueji, Vikas Yadav, Sathwik Tejaswi Madhusudhan
Arxiv preprint
Improving Selective Visual Question Answering by Learning from Your Peers
Corentin Dancette, Spencer Whitehead, Rishabh Maheshwary, Ramakrishna Vedantam, Stefan Scherer, Xinlei Chen, Matthieu Cord, Marcus Rohrbach
CVPR 2023
Practice Makes a Solver Perfect: Data Augmentation for Math Word Problem Solvers
Vivek Kumar, Rishabh Maheshwary, Vikram Pudi
NAACL 2022 (Oral)
A Strong Baseline for Query Efficient Attacks in a Black Box Setting
Rishabh Maheshwary*, Saket Maheshwary*, Vikram Pudi
EMNLP 2021
Adversarial Examples for Evaluating Math Word Problem Solvers
Vivek Kumar*, Rishabh Maheshwary*, Vikram Pudi
EMNLP 2021, Findings
Generating natural language attacks in a hard label black box setting
Rishabh Maheshwary, Saket Maheshwary, Vikram Pudi
AAAI 2021
A context aware approach for generating natural language attacks
Rishabh Maheshwary, Saket Maheshwary, Vikram Pudi
AAAI 2021, Student Poster