Aishwarya Agarwal
Research Associate, Adobe | PhD (Part-time), IIIT Hyderabad
I’m a Research Associate at Adobe Research India, and a part-time PhD student at CVIT, IIIT Hyderabad. I'm advised by Dr. Vineet Gandhi and co-advised by Dr. Srikrishna Karanam.
My research focuses on representation learning in low-resource settings, explainability in vision-language models (e.g., CLIP), and efficient adaptation and control of diffusion models for creative and task-agnostic image generation.
I hold a Dual Degree (BTech + MTech) from IIT Bombay in Electrical Engineering and AI & Data Science. I previously interned at Adobe under the guidance of Dr. Balaji Vasan Srinivasan, working on multimodal understanding and scene enrichment.
News
-
April 9, 2026Our work CCI accepted for presentation at the 5th Explainable AI for Computer Vision (XAI4CV) Workshop at CVPR 2026.
-
April 8, 2026GENIE, our framework for task-agnostic image transformation using diffusion models, got accepted at TMLR! This work was done in collaboration with Dr. Biplab Banerjee and his group at IIT Bombay.
-
February 21, 2026Two papers accepted at CVPR 2026 (one in the main track and one in findings). CCI presents striking visualization results for CLIP, while LiteEmbed enables adapting CLIP to rare classes using just a few images, without modifying the model.
-
October 15, 2025Our work on improving text rendering with diffusion models got accepted in NeurIPS UniReps Workshop 2025!
-
July 4, 2025Work on initial latent optimization for improved image geenration with diffusion models accepted to ACM MM 2025!
-
March 28, 2025Our work on enabling disentangled color-style control in diffusion models got accepted in CVPR CVEU Workshop 2025!
-
Feb 27, 2025Two papers accepted at CVPR 2025. Our paper TIDE on training locally interpretable models is selected as a Highlight!
-
Oct 29, 2024Two papers accepted at WACV 2025 on training-free diffusion model customization while balancing reconstruction-editability tradeoff.
-
Oct 23, 2023Our iterative image editing work using diffusion models is accepted to WACV 2024.
-
Jul 14, 2023Work on improving image-text alignment in diffusion models accepted to ICCV 2023!
-
Mar 4, 2023Sketchbuddy, our project on assisted sketching, is accepted to ACM MMSys 2023. Part of my undergrad internship at Adobe!
-
Aug 26, 2022Research on open-set cross-domain generalization accepted to WACV 2023 — part of my IITB thesis with Adobe.
-
Jul 12, 2022Joined Adobe Research, Bangalore!.
-
June 24, 2022Our work on few-shot class-incremental learning accepted to ACM Multimedia 2022. My first lead-author paper with Prof. Biplab!
Publications
2026
-
Concept Regions Matter: Benchmarking CLIP with a New Cluster-Importance Approach
In Computer Vision and Pattern Recognition (CVPR main), 2026 . -
LiteEmbed: Adapting CLIP to Rare Classes
In Computer Vision and Pattern Recognition (CVPR Findings), 2026 .
2025
-
Cocono: Attention contrast-and-complete for initial noise optimization in text-to-image synthesis
In ACM International Conference on Multimedia (ACM MM), 2025. -
Training-free Color-Style Disentanglement for Constrained Text-to-Image Synthesis
In Computer Vision and Pattern Recognition Workshop (CVPRW), 2025. -
TIDE: Training Locally Interpretable Domain Generalization Models Enables Test-time Correction
In Computer Vision and Pattern Recognition (CVPR), 2025 (Highlight). -
Composing Parts for Expressive Object Generation
In Computer Vision and Pattern Recognition (CVPR), 2025. -
An Image is Worth Multiple Words: Multi-attribute Inversion for Constrained Text-to-Image Synthesis
In Winter Conference on Applications of Computer Vision (WACV), 2025. -
AlignIT: Enhancing Prompt Alignment in Customization of Text-to-Image Models
In Winter Conference on Applications of Computer Vision (WACV), 2025.
2024
-
Iterative Multi-Granular Image Editing Using Diffusion Models
In Winter Conference on Applications of Computer Vision (WACV), 2024.
2023
-
ASTAR: Test-time Attention Segregation and Retention for Text-to-image Synthesis
In International Conference on Computer Vision (ICCV), 2023. -
SketchBuddy: Context-Aware Sketch Enrichment and Enhancement
In ACM Multimedia Systems Conference (ACM MMSys), 2023. -
Contrastive Learning of Semantic Concepts for Open-set Cross-domain Retrieval
In Winter Conference on Applications of Computer Vision (WACV), 2023.
2022
-
Semantics-Driven Generative Replay for Few-Shot Class Incremental Learning
In ACM International Conference on Multimedia (ACM MM), 2022.
2021
-
MIMOQA: Multimodal Input Multimodal Output Question Answering
In North American Chapter of the Association for Computational Linguistics (NAACL-HLT), 2021.
Email: agarwal.aishwarya2013@gmail.com