🎉Excited to share that our paper "Localizing Knowledge in Diffusion Transformers" has been accepted to NeurIPS 2025! 🌐Project Page: https://t.co/4PRKXy3rJv 📄Paper: https://t.co/UY0oPDhRTp Joint work with @BasuSamyadeep, @RezaeiKeivan, Zihao Lin, @nagsayan112358, @FeiziSoheil

🎉 Excited to share that our paper Localizing Knowledge in Diffusion Transformers has been accepted to #NeurIPS2025! In this work, we extend our previous study on localizing knowledge within T2I models to DiTs such as FLUX, PixArt, and SANA. paper: https://t.co/uouXmJWqx6

🚀New Paper: Localizing Knowledge in Diffusion Transformers 🌐Project page: https://t.co/NXFJm9Twkp 📄Paper: https://t.co/UY0oPDhRTp Joint work with: @BasuSamyadeep, @RezaeiKeivan, Zihao Lin, @nagsayan112358, @FeiziSoheil

🚨 We propose an actor-critic based framework that distills structured, step-by-step reasoning into AVLLMs at test time, improving their ability to process complex multi-modal inputs. 📢 We present AVReasonBench, a challenging benchmark comprising 4500 AV questions.

Note that: - You must be a PhD student at the time of internship. - Students with publication(s) at tier-1 conferences are preferred. - The location is in India.

🚀 Internship Opportunity at #AdobeResearch🚀 Looking for PhD interns for Summer 2025! Interested in exploring the intersection of multimodal LLMs, diffusion models, etc? 📩 Send me a DM with your CV, website, and GScholar profile. #GenerativeAI

📢📢 We are thrilled to share that our paper Meerkat : Audio-Visual Large Language Model for Grounding in Space and Time got accepted to @eccvconf 2024! 🚀🎊 🧵 (1/n)

Our work "SAFARI" got accepted at ECCV 2024. We introduce a novel method for solving RES task..we introduce a novel of doing layer level fusion between text and images. Wonderful work done by @nagsayan112358 #ECCV2024 #multimodal Preprint coming soon

📢 Excited to share that SAFARI got accepted to ECCV 2024! 🎉 We develop a novel method for solving Referring Expression Segmentation (RES) task under low annotation settings. Big congrats to my coauthors @koustavagoswami, Srikrishna Karanam. #ECCV2024 #multimodal #visionlanguage

@schowdhury671 will be presenting MeLFusion and @Shramanpramani2 will be presenting VistaLLM

Unfortunately won't be attending #CVPR24 but please check out our papers 1. MeLFusion: https://t.co/9NKE4Ys3ch 2. VistaLLM: https://t.co/VNn1Hb1aYo Both of these are Poster Highlights! #CVPR24 #GenerativeAI

Also in session #6, @schowdhury671 presents "MeLFusion: Synthesizing Music from Image and Language Cues using Diffusion Models" ( https://t.co/2I9j0ysfAG) with @nagsayan112358 @josephkj_in @balajivasan @dmanocha @AdobeResearch

For the first time, we show that the Llama 7B LLM can be trained on a single consumer-grade GPU (RTX 4090) with only 24GB memory. This represents more than 82.5% reduction in memory for storing optimizer states during training. Training LLMs from scratch currently requires huge

📢Happy to share that EgoVLPv2 (ICCV 2023) is awarded as an Egocentric Vision 2022/2023 Distinguished Paper. Thanks to my coauthors @yalesong @PengchuanZ @MikeShou1 @nagsayan112358 @KevinQHLin arXiv: https://t.co/3OlhdwV0XR Code: https://t.co/VqjFhp4Stn… @CVPR @ICCVConference

@WenhuChen Nice work! Reasoning over multiple images is indeed a less-explored field. We have reported results on NLVR and Co-Segmentation tasks using our proposed VistaLLM in CVPR 2024.

#CVPR2024 has said an ‘Accept’ to our work on generating music conditioned on multimodal signals from text and images! Joint work with our 2023 @AdobeResearch interns (@schowdhury671 and @nagsayan112358) and @balajivasan.

📢 Thrilled to share that our paper MeLFusion: Synthesizing Music from Image and Language Cues using Diffusion Models got accepted to CVPR 2024! In this work we present a novel diffusion powered approach to generate audio from multi-modal inputs (image and text).

@schowdhury671 APoLLo substantially improves the generalization capabilities of VLP models when fine-tuned in a few-shot setting. We introduce trainable cross-attention-based adapter layers in conjunction with vision and language encoders to strengthen the alignment between the two modalities.

We are happy to share that we recently presented our work APoLLo🚀: Unified Adapter and Prompt Learning for Vision Language Models at #EMNLP2023 (Main track). Huge thanks to @schowdhury671 and Dinesh Manocha. Paper:

The code and model checkpoints of VoLTA (TMLR, 2023) are now publicly available! 📜 Paper: https://t.co/pfCc0SVIDz 📷 Code: https://t.co/U2Il7DZqdd 📷Project: https://t.co/2h1HOBOYvP Huge thanks to @Shramanpramani2, Li Jing, @JiachenAI, Hardik Shah, @ylecun, and Rama Chellappa.