🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Thrilled to share some of our latest work on breaking down reasoning for table understanding. Amazing work led by Zilong (Ryan) Wang and will be presented in ICLR!

Stop by the #ICLR2024 Google booth today at 9:30 AM to hear Zilong (Ryan) Wang, Lesly Miculicich, and Chen-Yu Lee describe how Chain-of-Table enables step-by-step reasoning from table inputs by simplifying complex tables into more informative and manageable segments.

We release PaliGemma. I'll keep it short, still on vacation: - sota open base VLM designed to transfer quickly, easily, and strongly to a wide range of tasks - Also does detection and segmentation - We provide lots of examples - Meaty tech report later! ai.google.dev/gemma/docs/pal…

Very excited about the open source release of the PaliGemma models. It was honor to work with amazing collaborators to make it available wherever you get your podcasts 🪇 VertexAI: console.cloud.google.com/vertex-ai/publ… HF: hf.co/spaces/google/… Fine-tuning Colab: colab.research.google.com/github/google-…

I'll join Google DeepMind in Zürich to work with Julian Eisenschlos, starting in June!! This has been a dream of mine since I've started working on NLP some 7 years ago. It's really great to have it come to fruition.

We are very thrilled to work with you this summer Benjamin Minixhofer!

It was a great pleasure to work with Mubashara Akhtar over the last few months, and I am happy to share the results of that effort: TANQ, the first dataset to benchmark if LLMs are able to not only find open domain answers but organize them in a structured form such as tables.

📡We have great news: Khipu 2025 already has a date! Can you guess where it will be ⁉️

Today at the 4p poster session in #ACL2024NLP I'll be presenting the awesome work led by Syrine Krichene, in collaboration with Francesco Piccinno and Fangyu Liu showing how to measure (and improve) faithfulness when summarising the content of charts arxiv.org/abs/2405.19094. Sawadee ka!🇹🇭

I'll be speaking next Tuesday in Zürich about our work on visual language understanding to go from pixels to structured knowledge zurichai.ch/events/zurichc…. Hope to see you there!

Our ZurichCV Meetup #5 will be on the 17th of September 18.00 at the ETH AI center. We have two amazing talks:  - Anton Obukhov (ETH Zurich) about Repurposing Text-to-Image Models.  - Julian Eisenschlos (Google Deepmind) about Visual language. RSVP: zurichai.ch/events/zurichc…

This is a cool feature I collaborated with, and I personally use quite a bit. Pretty excited for it to be out there for everyone to try out!

I did the thing: eisenjulian.bsky.social

Stoked to get this model out in the wild. I had the pleasure to work with super talented colleagues for more than a year to improve how our models interact with and generate structured images. Up and to the right 📈

Online now: Ecosystem ecology needs an ecology of molecules dlvr.it/THlbBr

Last week over in khipu.ai 🇨🇱was a fantastic reminder of the value of community and it left me feeling super optimistic about the future of AI in LatAm if we build it together. Thank you to the rest of the organizing crew for making magic happen once more. Can't wait for 27!

What is Khipu for you? For me it sounds like family 🧡 #khipu2025 khipu.ai

🚨 New Benchmark Alert Our popular ChartQA📊benchmark is now saturated⚠️ Recent ones like CharXiv fall short in visual & question diversity. 🔥 Say hello to ChartQAPro: A more diverse & challenging benchmark for Chart Question Answering! 🧵👇 📄 arxiv.org/abs/2504.05506…