A framework for generating and reversing physical actions in vision-language models, probing causal and temporal reasoning capabilities.
@article{kadambi2025doundo,title={Do-Undo: Generating and Reversing Physical Actions in Vision-Language Models},author={Mahajan, Shweta and Kadambi, Shreya and Le, Hoang and Hayat, Munawar and Porikli, Fatih},journal={arXiv},year={2025},url={https://doi.org/10.48550/arXiv.2512.13609},}
ICCV-W
MADI: Masking-Augmented Diffusion with Inference-Time Scaling for Visual Editing
Shreya Kadambi, Risheek Garrepalli, Shubhankar Borse, and 2 more authors
A masking-augmented diffusion framework with inference-time scaling for visual editing tasks.
@article{kadambi2025madi,title={MADI: Masking-Augmented Diffusion with Inference-Time Scaling for Visual Editing},author={Kadambi, Shreya and Garrepalli, Risheek and Borse, Shubhankar and Hayat, Munawar and Porikli, Fatih},journal={arXiv},year={2025},note={ICCV Workshop, 2025},url={https://doi.org/10.48550/arXiv.2507.13401},}
arXiv
SubZero: Composing Subject, Style, and Action via Zero-Shot Personalization
Shubhankar Borse, Kartikeya Bhardwaj, Mohammad Reza Karimi Dastjerdi, and 8 more authors
Zero-shot personalization for diffusion models enabling compositional control over subject, style, and action simultaneously.
@article{borse2025subzero,title={SubZero: Composing Subject, Style, and Action via Zero-Shot Personalization},author={Borse, Shubhankar and Bhardwaj, Kartikeya and Karimi Dastjerdi, Mohammad Reza and Park, Hyojin and Kadambi, Shreya and Shivakumar, Shobitha and Mandke, Prathamesh and Nayak, Ankita and Teague, Harris and Hayat, Munawar and Porikli, Fatih},journal={arXiv},year={2025},url={https://doi.org/10.48550/arXiv.2502.19673},}
arXiv
DuoLoRA: Cycle-consistent and Rank-disentangled Content-Style Personalization
Aniket Roy, Shubhankar Borse, Shreya Kadambi, and 8 more authors
A cycle-consistent LoRA approach for disentangled content and style personalization in generative models.
@article{roy2025duolora,title={DuoLoRA: Cycle-consistent and Rank-disentangled Content-Style Personalization},author={Roy, Aniket and Borse, Shubhankar and Kadambi, Shreya and Das, Debasmit and Mahajan, Shweta and Garrepalli, Risheek and Park, Hyojin and Nayak, Ankita and Chellappa, Rama and Hayat, Munawar and Porikli, Fatih},journal={arXiv},year={2025},url={https://doi.org/10.48550/arXiv.2504.13206},}
arXiv
MultiHuman-Testbench: Benchmarking Image Generation for Multiple Humans
Shubhankar Borse, Seokeon Choi, Sunghyun Park, and 6 more authors
A benchmark for evaluating multi-human image generation across face count, identity similarity, prompt alignment, and action detection metrics.
@article{borse2025multihuman,title={MultiHuman-Testbench: Benchmarking Image Generation for Multiple Humans},author={Borse, Shubhankar and Choi, Seokeon and Park, Sunghyun and Kim, Jeongho and Kadambi, Shreya and Garrepalli, Risheek and Yun, Sungrack and Hayat, Munawar and Porikli, Fatih},journal={arXiv},year={2025},url={https://doi.org/10.48550/arXiv.2506.20879},}
2024
NeurIPS
FouRA: Fourier Low-Rank Adaptation
Shubhankar Borese, Shreya Kadambi, Nilesh Prasad Pandey, and 7 more authors
In Advances in Neural Information Processing Systems, 2024
Fourier-domain low-rank adaptation for efficient fine-tuning of large vision models, improving expressivity over standard LoRA while maintaining parameter efficiency.
@inproceedings{kadambi2024foura,title={FouRA: Fourier Low-Rank Adaptation},author={Borese, Shubhankar and Kadambi, Shreya and Pandey, Nilesh Prasad and Bhardwaj, Kartikeya and Ganapathy, Viswanath and Priyadarshi, Sweta and Garrepalli, Risheek and Esteves, Rafael and Hayat, Munawar and Porikli, Fatih},booktitle={Advances in Neural Information Processing Systems},year={2024},url={http://papers.nips.cc/paper_files/paper/2024/hash/83960718b4d12f799985206f1b1cf00f-Abstract-Conference.html},}
NeurIPS
Sparse High Rank Adapters
Kartikeya Bhardwaj, Nilesh Prasad Pandey, Sweta Priyadarshi, and 9 more authors
In Advances in Neural Information Processing Systems, 2024
A sparse high-rank adapter paradigm enabling rapid adapter switching and multi-adapter fusion with no inference overhead, outperforming LoRA in both efficiency and concept preservation.
@inproceedings{bhardwaj2024shira,title={Sparse High Rank Adapters},author={Bhardwaj, Kartikeya and Pandey, Nilesh Prasad and Priyadarshi, Sweta and Ganapathy, Viswanath and Kadambi, Shreya and Esteves, Rafael and Borse, Shubhankar and Whatmough, Paul N. and Garrepalli, Risheek and van Baalen, Mart and Teague, Harris and Nagel, Markus},booktitle={Advances in Neural Information Processing Systems},year={2024},url={http://papers.nips.cc/paper_files/paper/2024/hash/18c0102cb7f1a02c14f0929089b2e576-Abstract-Conference.html},}
2023
GLOBECOM
Neural 5G Indoor Localization with IMU Supervision
Aleksandr Ermolov, Shreya Kadambi, Maximilian Arnold, and 8 more authors
In IEEE Global Communications Conference (GLOBECOM), 2023
Neural network approach for joint 5G indoor localization leveraging IMU supervision to improve positioning accuracy without ground-truth location labels.
@inproceedings{ermolov2023neural5g,title={Neural 5G Indoor Localization with IMU Supervision},author={Ermolov, Aleksandr and Kadambi, Shreya and Arnold, Maximilian and Hirzallah, Mohammed and Amiri, Roohollah and Singh, Deepak and Yerramalli, Srinivas and Dijkman, Daniel and Porikli, Fatih and Yoo, Taesang and Major, Bence},booktitle={IEEE Global Communications Conference (GLOBECOM)},year={2023},url={https://doi.org/10.48550/arXiv.2402.09948},}
ICLR
WiNeRT: Towards Neural Ray Tracing for Wireless Channel Modelling and Differentiable Simulations
Tribhuvanesh Orekondy, Kumar Pratik, Shreya Kadambi, and 3 more authors
In International Conference on Learning Representations (ICLR), 2023
NeRF-inspired neural surrogate for differentiable wireless electromagnetic propagation modelling in indoor environments.
@inproceedings{orekondy2023winert,title={WiNeRT: Towards Neural Ray Tracing for Wireless Channel Modelling and Differentiable Simulations},author={Orekondy, Tribhuvanesh and Pratik, Kumar and Kadambi, Shreya and Ye, Hao and Soriaga, Joseph and Behboodi, Arash},booktitle={International Conference on Learning Representations (ICLR)},year={2023},url={https://openreview.net/forum?id=tPKKXeW33YU},}
2022
ICC
Neural RF SLAM for Unsupervised Positioning and Mapping with Channel State Information
Shreya Kadambi, Arash Behboodi, Joseph B. Soriaga, and 4 more authors
In IEEE International Conference on Communications (ICC), 2022
An unsupervised neural network architecture for joint user localization and environment mapping from channel state information, using a physics-based encoder-decoder with virtual anchors.
@inproceedings{kadambi2022neuralrfslam,title={Neural RF SLAM for Unsupervised Positioning and Mapping with Channel State Information},author={Kadambi, Shreya and Behboodi, Arash and Soriaga, Joseph B. and Welling, Max and Amiri, Roohollah and Yerramalli, Srinivas and Yoo, Taesang},booktitle={IEEE International Conference on Communications (ICC)},year={2022},url={https://arxiv.org/abs/2203.09513},}
2021
WGAN
WGAN Domain Adaptation for the Joint Optic Disc-and-Cup Segmentation in Fundus Images
Domain adaptation using Wasserstein GANs for joint optic disc and cup segmentation in retinal fundus images.
@article{kadambi2021wgan,title={WGAN Domain Adaptation for the Joint Optic Disc-and-Cup Segmentation in Fundus Images},author={Kadambi, Shreya and Wang, Zeya and Xing, Eric},year={2021},}