🔥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

Had a good breakfast today!

It can be helpful to pinpoint the in-context information that a language model uses when generating content (is it using provided documents? or its own intermediate thoughts?). We present Attribution with Attention (AT2), a method for doing so efficiently and reliably! (1/8)

We are at ICLR this week! If you are interested in OpenAI, our team, or just generally chatting about AI, please reach out!

This work has been a long time coming and I'm so grateful to my collaborators for helping make this work possible. Main takeaway: AI supply chains matter! We've seen them emerge (rapidly) in the past few years and they will have implications on *all* of us, inside and outside

AI supply chains introduce new challenges to traditional formulations of deployment, safety, & evaluation. But! They can be studied! Come join the conversation (& read our new paper ✨) !

Ever wondered which data from large datasets (like OXE) actually helps when training/tuning a policy for specific tasks? We present DataMIL, a framework for measuring how each training sample influences policy performance, hence enabling effective data selection 🧵

Modern reasoning models think in plain English. Monitoring their thoughts could be a powerful, yet fragile, tool for overseeing future AI systems. I and researchers across many organizations think we should work to evaluate, preserve, and even improve CoT monitorability.