Publications

CRAFT: Curriculum Rank Adversarial Fine-Tuning for Robust Vision Language Models

In this work, we propose CRAFT, a unified low-rank fine-tuning framework designed to enhance both out-of-distribution (OOD) and adversarial robustness by integrating adaptive adversarial weight perturbations into a curriculum-driven Low-Rank Adaptation (LoRA) framework.

Shivang Chopra, Chengyue Huang, Brisa Maneechotesuwan, Zsolt Kira.

CRAFT: Curriculum Rank Adversarial Fine-Tuning for Robust Vision Language Models

FRAMES-VQA: Benchmarking Fine-Tuning Robustness across Multi-Modal Shifts in Visual Question Answering

In this work, we propose a new benchmark FRAMES-VQA (Fine-Tuning Robustness Across Multi-Modal Shifts in VQA) for evaluating robust fine-tuning for VQA tasks.

Chengyue Huang, Brisa Maneechotesuwan, Shivang Chopra, Zsolt Kira.

FRAMES-VQA: Benchmarking Fine-Tuning Robustness across Multi-Modal Shifts in Visual Question Answering

Directional Gradient Projection for Robust Fine-tuning of Foundation Models

In this work, we propose Drectional Grdient Projection (DiGraP), a novel layer-wise trainable method that incorporates directional information from gradients to improve OOD robustness.

Chengyue Huang, Junjiao Tian, Brisa Maneechotesuwan, Shivang Chopra, Zsolt Kira.

Refining Text-to-Image Generation: Towards Accurate Training-Free Glyph-Enhanced Image Generation

In this work, we introduce a training-free framework to enhance the capability of Diffusion Models to generate visual text.

Sanyam Lakhanpal, Shivang Chopra, Vinija Jain, Aman Chadha, Man Luo

Source-Free Domain Adaptation with Diffusion-Guided Source Data Generation

This paper introduces a novel approach to leverage the generalizability capability of Diffusion Models for Source-Free Domain Adaptation (DM-SFDA).

Shivang Chopra, Suraj Kothawad, Houda Aynaou, Aman Chadha

Source-Free Domain Adaptation with Diffusion-Guided Source Data Generation

HyWay: Enabling Mingling in the Hybrid World

We present HyWay, short for “Hybrid Hallway”, to enable mingling and informal interactions among physical and virtual users, in casual spaces and settings, such as office water cooler areas, conference hallways, trade show floors, and more.

Harsh Vijay, Saumay Pushp, Amish Mittal, Praveen Gupta, Meghna Gupta, Sirish Gambhira, Shivang Chopra, Mayank Baranwal, Arshia Arya, Ajay Manchepalli, Venkata N Padmanabhan

HyWay: Enabling Mingling in the Hybrid World

Symbiotic Artificial Intelligence: Order Picking And Ambient Sensing

Using egocentric video and head motion data from 67 order picking tasks (244 picks;149 orders), we learn visual models of the 10 objects picked to fulfill the orders.

Zhe Ming Chng, Calix Tang, Darshan Krishnaswamy, Haoyang Yang, Shivang Chopra, Jon Womack, Thad Starner

Learning to Discern: Imitating Heterogeneous Human Demonstrations with Preference and Representation Learning

In this work, we introduce Learning to Discern (L2D), an offline imitation learning framework for learning from demonstrations with diverse quality and style.

Sachit Kuhar, Shuo Cheng, Shivang Chopra, Matthew Bronars, Danfei Xu

Learning to Discern: Imitating Heterogeneous Human Demonstrations with Preference and Representation Learning

Active Data Discovery: Mining Unknown Data using Submodular Information Measures

Active Learning is a very common yet powerful framework for iteratively and adaptively sampling subsets of the unlabeled sets with a …

Suraj Kothawad, Shivang Chopra, Saikat Ghosh, Rishabh Iyer

Active Data Discovery: Mining Unknown Data using Submodular Information Measures

Attention-Net: An Ensemble Sketch Recognition Approach Using Vector Images

For the past few decades, machines have replaced humans in several disciplines. However, machine cognition still lags behind the human …

Gaurav Jain, Shivang Chopra, Suransh Chopra, Anil Singh Parihar

Hindi-english hate speech detection: Author profiling, debiasing, and practical perspectives

Code-switching in linguistically diverse, low resource languages is often semantically complex and lacks sophisticated methodologies …

Shivang Chopra, Ramit Sawhney, Puneet Mathur, Rajiv Ratn Shah

Hindi-english hate speech detection: Author profiling, debiasing, and practical perspectives

Utilizing temporal psycholinguistic cues for suicidal intent estimation

Temporal psycholinguistics can play a crucial role in studying expressions of suicidal intent on social media. Current methods are …

Puneet Mathur, Ramit Sawhney, Shivang Chopra, Maitree Leekha, Rajiv Ratn Shah

Utilizing temporal psycholinguistic cues for suicidal intent estimation

Open Domain Suggestion Mining Leveraging Fine-Grained Analysis (Workshop Paper)

In this work we propose a two-tier pipeline that leverages Discourse Marker based oversampling and fine-grained suggestion mining techniques to retrieve suggestions from online forums.

Shreya Singal, Tanishq Goel, Shivang Chopra, Sonika Dahiya

Open Domain Suggestion Mining Leveraging Fine-Grained Analysis (Workshop Paper)