One of the coolest features of word embeddings is to perform arithmetic. For example, woman + (king - man) = queen. Is it be possible to do the same visually? In our new #CVPR2022 paper, we show that in CLIP's embedding space, it is. Code: [1/3]

RT @AnthropicAI: Today we're releasing Claude Opus 4.1, an upgrade to Claude Opus 4 on agentic tasks, real-world coding, and reasoning. htt….

RT @OpenAI: Our open models are here. Both of them.

RT @itai_gat: Excited to share our recent work on corrector sampling in language models! A new sampling method that mitigates error accumul….

RT @ComscienceBiu: Prof. @GalChechik & Dr. @idansc for organized and led the symposium on multimodal deep learning at @BarIlanU’s CS Dept….

RT @YftahZ: 1/6🚀 New #ACL2025Findings:.We show you can predict if Chain-of-Thought (CoT) reasoning will succeed — before any tokens are gen….

RT @ItamarZimerman: 📄🚨New!.Attribution methods that assign relevance to tokens are key to extracting explanations from Transformers and LLM….

RT @itayhzn: 🧵1/.Text-to-video models generate stunning visuals, but… motion? Not so much. You get extra limbs, objects popping in and out.….

RT @_akhaliq: RefVNLI. Towards Scalable Evaluation of Subject-driven Text-to-image Generation

RT @Bar_ilan: 🔬 יום פתוח לתארים מתקדמים בבר-אילן!.📅 יום רביעי 2.4 | 15:00-18:00 | בניין ננוטכנולוגיה.פגישות אישיות עם חוקרים מובילים ב-STEM….

RT @yairshp: 🔎Unlike methods like DreamBooth (which rely on MSE and fail with a single image), SISO leverages pre-trained similarity metric….

RT @yairshp: 🚀Introducing SISO – a plug-and-play approach for image personalization using just one image!

RT @guy_yariv: I'm thrilled to announce that Through-The-Mask (TTM) has been accepted to #CVPR2025! . TTM is an I2V generation framework th….

RT @hila_chefer: VideoJAM is our new framework for improved motion generation from @AIatMeta. We show that video generators struggle with m….

RT @YftahZ: Excited to share our new EMNLP paper! 📄 We uncover that different LLM layers naturally play distinct roles in multitask learnin….

RT @SchmidhuberAI: The #NobelPrizeinPhysics2024 for Hopfield & Hinton rewards plagiarism and incorrect attribution in computer science. It'….

RT @hila_chefer: Introducing✨Still-Moving✨—our work from @GoogleDeepMind that lets you apply *any* image customization method to video mode….

And if inference time is a concern, by optimizing a token, we can easily reuse it with different prompts without any need for iterative optimization. A joint work with Lior Wolf. Congratulations to the leading author, Oz Zafar. [3/3]

We optimize token embeddings using a derivable estimator that aggregate an object potential. The token is trained iteratively generating images at each optimization step. However, using object potential is not accurate since different viewpoints can lead to different potential

Ever tried creating an image with exactly N objects? It’s nearly impossible with current text-to-image methods. In our latest work, we address this challenge and introduce a new iterative counting optimization for SD. Try our colab: Project page:

RT @itai_gat: Excited to share Discrete Flow Matching! A discrete flow framework that yields state-of-the-art non-autoregressive modeling.….