🔥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

Wow, generative evals show K2 not Grok 4 is what you should be hyped about 😮 Come chat with me about how you can convert your MCQ dataset to Generative evals ICML Conference. Also presenting this work at the (World Models) Metrics for Evaluating Understanding Workshop on Friday

Excited to be heading to ICML this year to present two projects, both as spotlights! 🎉 Big thanks to my collaborators — come say hi if you're around! #ICML2025 #ML

Combine the best parts of SF and Vancouver and you'd probably get paradise

Presenting today at #ICML2025. To learn how to measure language model similarity, and it's effects on LLM as a Judge and Weak to Strong distillation, join our poster session: Today 11 am -1:30 pm, East Exhibition Hall A-B E-2411 w/ Ameya P. @ ICML 2025 joschkastrueber Ilze Amanda Auzina

Flashing research endorsement is a bit weird... promotes ad hominem hero worship, taking away the focus from the science

Come and find me today at #ICML2025 and let's talk about speech 💬 decoding from the brain and scaling brain-computer interfaces 🤖. 11 am -1:30 pm, West Exhibition Hall, Poster W-415

Have a peek at our early work on non-invasive brain-to-text 🧠💬 where we decode phrases and sentences directly from the brain. Poster up all day tomorrow 👀 at the GenBio Workshop @ ICML25 at #ICML2025. arxiv.org/abs/2505.13446

Just today presented the bronze medal market (at 86%, coincidentally) to people at an ICML poster session on forecasting evaluations... Crazy OpenAI just shot past all the intermediate medals to gold. Next research wave will be about figuring out the new technique...

I feel this may be helpful to some of you today:

ICML'25 is a wrap. Heads down for ICLR deadline now

One weird things about harder and harder benchmarks is that validating these questions becomes even harder. Efforts like this that still do it are extremely valuable

The gap between big frontier api models and small efficient ones is much larger than benchmarks reflect