Our research codebase for Diffusion-DPO is now public! github.com/SalesforceAIRe… This is the exact (cleaned up) code we used for the primary results in our paper. It also has an SDXL-Turbo script which works well (see x.com/bram_wallace/s…). Enjoy!

Check out our recent #TMLR2024 paper (led by Simar Kareer) with a rather surprising finding: existing video adaptation methods don't actually use videos _effectively_⁉️ Bonus: We open source a unified library (built on top of mmseg) to faciliate future Video DA research Machine Learning at Georgia Tech

If you're at #ICLR2024🇦🇹, stop by to hear Andrew Liao present his exciting work on characterizing and addressing **annotation** mismatches across datasets!🔥

Looking forward to #CVPR2024 this week, on home turf for a change! DM if you’re around and want to meet / catch up!

Evaluate the hallucination of your VLMs using our new benchmark

Introducing EgoMimic - just wear a pair of Project Aria Project Aria @Meta smart glasses 👓 to scale up your imitation learning datasets! Check out what our robot can do. A thread below👇

🤔Ever wondered why merging LoRA models is trickier than fully-finetuned ones? 🔍We explore this and discover that poor alignment b/w LoRA models lead to subpar merging. 💡The solution? KnOTS🪢— our latest work that uses SVD to improve alignment and boosts SOTA merging methods.

Looking forward to some Miami sun this week at #EMNLP2024, my first NLP conference in ~7 years! ☀️ HMU if you’d like to learn more about our work at Salesforce AI Research or just meet/catch up! 🍹

Introducing Gaze-LLE, a new model for gaze target estimation built on top of a frozen visual foundation model! Gaze-LLE achieves SOTA results on multiple benchmarks while learning minimal parameters, and shows strong generalization paper: arxiv.org/abs/2412.09586

Join us at the first-ever EMACS workshop #CVPR2025! 🚨 Submissions open March 5: tinyurl.com/emacs25 See you in Nashville! 🎸 #CVPR2025

🚀 Excited about how generative AI can power experimental (not just observational) audits of ML systems that reveal actionable insights into performance and bias? Join us at EMACS (Experimental Model Auditing with Controllable Synthesis) workshop #CVPR2025! sites.google.com/view/emacs2025/

🚨🚨 Paper submission deadline extended to May 4. Submit your work (in-progress or complete!) to the EMACS workshop @CVPR2025 in Nashville! Submission link: tinyurl.com/emacs2025submit #CVPR2025 #GenerativeAI #bias

Happening now in 208B, come check out the first EMACS workshop! #CVPR2025

🚀 Computer-using agents represent a powerful new paradigm for human-computer interaction. Over the past year, we’ve explored multiple approaches to tackle the key challenges in building robust CUA systems. 12/2024 we released Aguvis (x.com/CaimingXiong/s…) 07/2024 we released

Thank you so much Caiming! We show that involving coding as a new type of action apart from GUI action for CUA can significantly help improve the computer-using performance while reducing the total actions for task solving. If you are interested in it, please take a look at