Strong picture!

Thrilled to see this result, showing our dedication to align Llama to not only be helpful but also capable of understanding safety nuances, with strong generalization to unseen external benchmarks!

Eagle-eyed readers might have spotted 🌈 Rainbow Teaming among the key prompt generation techniques used for Llama 3 This was always the North Star for our project, and I'm thrilled that Mikayel Samvelyan, Sharath Raparthy and I pulled it off! Here's what we learned along the way 🧵

Today, we release several Moshi artifacts: a long technical report with all the details behind our model, weights for Moshi and its Mimi codec, along with streaming inference code in Pytorch, Rust and MLX. More details below 🧵 ⬇️ Paper: kyutai.org/Moshi.pdf Repo:

🚨We are hiring PhD Interns at Meta Ai for summer 2025! 🚨 Broadly, my team is working on understanding and improving the intersection between Ai and Society (social systems). We have an excellent team of researchers excited to have you! … 🧵1/3

This seems like really cool work, I'm excited to give it a read! It's awesome that our paper on interpretable jailbreaks, led by @littlefish3625, can be leveraged to make models more robust. This seems like a nice case study in the real world applications of interpretability!

Want to know if a ML model was trained on your dataset? Introducing ✨Data Taggants✨! We use data poisoning to leave a harmless and stealthy signature on your dataset that radiates through trained models. Learn how to protect your dataset from unauthorized use... A 🧵

My twitter feed this weekend is full of people praising Roberta and recommending this internship opportunity 😂 I agree so I’m adding one more tweet to this list!

Job alert 🚨 My team AI at Meta is looking for a PhD intern to join us in 2025 in Paris. We are working on self-supervised learning from video, world modelling and JEPA ! Apply here or reach out directly: metacareers.com/jobs/168411027…

5-star opportunity!!! 💫 I 100% recommend Grégoire Mialon, Dheeraj Mekala and the food

Super excited to share 🧠MLGym 🦾 – the first Gym environment for AI Research Agents 🤖🔬 We introduce MLGym and MLGym-Bench, a new framework and benchmark for evaluating and developing LLM agents on AI research tasks. The key contributions of our work are: 🕹️ Enables the

I am at #ICLR and honored to present this work on Saturday afternoon at the poster session. Thanks Lei Yu Karen Hambardzumyan Nicola Cancedda for this wonderful collaboration! I am also happy to chat about Llama / agents / safety 👋

Very cool work from Meta Superintelligence Lab. They are open-sourcing Meta Agents Research Environments (ARE), the platform they use to create and scale agent environments. Great resource to stress-test agents in environments closer to real apps. Read on for more:

🧠Great research from Meta Superintelligence Labs. Proposes Meta Agents Research Environments (ARE) for scaling up agent environments and evaluations. ARE lets researchers build realistic agent environments, run agents asynchronously, and verify them cleanly. On top of it

We need better agent evaluations! Glad to have collaborated with Meta Super Intelligence Lab to release Gaia2 and ARE! GPT5 (high) from OpenAI is leading on execution, search, ambiguity, adaptability and noise. Kimi-K2 from Kimi.ai is leading open weight. Full

Gaia2 Leaderboard Update: DeepSeek is leading OSS models! We’ve added fresh models (DeepSeek v3.1, Qwen 235B thinking, GPT-OSS 120B) and uncovered cool insights on cost, reasoning, and efficiency. Blogpost 👉 tinyurl.com/ybvxtmny