📢 WMT24 Metric Shared Task Starts NOW! 🎉 Calling all researchers and metrics enthusiasts! We invite you to submit your best evaluation metrics to the WMT Metric Shared Task. Assign quality scores to all WMT submissions and translations by top LLMs like Gemini, GPT4o, Claude,

We're cooking something new in TRL with DPO and online feedback from reward models 🧑‍🍳 The neat thing here is that you sample from the model during training to keep the distribution on-policy and improve overall performance -- like RLHF, but without the pain 😅 Early experiments

Happy to share that the online DPO work is now in TRL! I was shocked that it worked so well, matching RLOO / PPO's performance in TL;DR 🔥Inspired by @ShawnGuo13, Michael Noukhovitch prototyped an online DPO codebase, and Quentin Gallouédec, Edward Beeching Lewis Tunstall and I brought it to TRL!

TRL v0.10.1 is here and it's beefy 💪 🔁 Online DPO by Google DeepMind for aligning better LLMs 🐯 Liger kernel integration from LinkedIn to supercharge SFT 🖼️ DPO for VLMs: 🌋 LLaVa, ✨ PaliGemma, 🐶 Idefics2 👩‍⚖️ Use LLMs as a judge for to compute win rates during training 🔍

🪩The State of AI 2024 has landed! 🪩 Our seventh installment is our biggest and most comprehensive yet, covering everything you *need* to know about research, industry, safety and politics. As ever, here's my director’s cut (+ video tutorial!) 🧵

Interested in doing research on Google Translate and Gemini? Good news! I’m hiring for full-time roles on the Google Translate Research Team! Apply here: google.com/about/careers/…

🧏 Learn how AI is revolutionizing sign language recognition. On this week's People of AI, ThadStarner & Sam Sepah discuss their work on sign language recognition & AI-powered accessibility tech. 👀 Watch → goo.gle/3YBGhv1 🎧 Listen → goo.gle/4hkP2AV

🚀 Join the Gemini Multilinguality team Google DeepMind 🌐 We’re looking for researchers passionate about making LLMs helpful for all. Dramatically improve model quality, coverage, and cultural relevance across hundreds of languages. #NLProc #MultilingualAI #i18n #LLMs

Franklin Goodman Cheng Lu NeurIPS Conference Fei-Fei Li Jiahui Yu Ed H. Chi I didn't see the talk, but the images I've seen of the slide seem quite offensive. Such generalizations should have no place in NeurIPS or anywhere else.

feeling a bit under the weather this week … thus an increased level of activity on social media and blog: kyunghyuncho.me/i-sensed-anxie…

docs.google.com/document/d/1aE…

Update: Felix's farewell letter was just published. I read each and every word. Sharing here with my big audience because it may literally save lives. Turns out that stress in AI only contributed a bit. Overdose of ketamine (psychedelic drug) was what pushed him over the edge.

Making LLMs run efficiently can feel scary, but scaling isn’t magic, it’s math! We wanted to demystify the “systems view” of LLMs and wrote a little textbook called “How To Scale Your Model” which we’re releasing today. 1/n

🚨New machine translation dataset alert! 🚨We expanded the language coverage of WMT24 from 9 to 55 en->xx language pairs by collecting new reference translations for 46 languages in a dataset called WMT24++ Paper: arxiv.org/abs/2502.12404… Data: huggingface.co/datasets/googl…

Download the model weights on Hugging Face and Kaggle to get started. We can't wait to see what you build with T5Gemma. huggingface.co/collections/go…

Introducing T5Gemma: the next generation of encoder-decoder/T5 models! 🔧Decoder models adapted to be encoder-decoder 🔥32 models with different combinations 🤗Available in Hugging Face and Kaggle developers.googleblog.com/en/t5gemma

🤖 Now on #KaggleModels! Learn more: kaggle.com/models/google/…

Voxtral uses online DPO!

Our Google Translate team is bringing a strong presence to #ACL2025 in Vienna this week! 🇦🇹 My group is excited to present several of our latest papers. 👇 Don't miss them!