🔥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

Scaling Test-time Compute (12 Minute Version)

???

the most underused, underappreciated, and underrated trick that is basically only used by Anthropic; context distillation

Sleep-time Compute Beyond Inference Scaling at Test-time

Looking forward to seeing folks at ICLR next week!

We're excited to release our latest paper, “Sleep-time Compute: Beyond Inference Scaling at Test-Time”, a collaboration with Charlie Snell from UC Berkeley and Letta advisors / UC Berkeley faculty Ion Stoica and Joey Gonzalez letta.com/blog/sleep-tim…

We explore a new dimension in scaling reasoning models in Adaptive Parallel Reasoning APR lets LMs learn to orchestrate both serial & parallel compute E2E via supervised training + RL — w/ better efficiency and scalability than long CoT on Countdown 🧵 arxiv.org/abs/2504.15466

woot woot 🎉

If you’re at ICLR and wondering why I just gave my talk in sweatpants this morning, it’s a long story, but it has something to do with flying in at 1AM last night and my luggage not making it to Singapore

Last day of PhD! I pioneered using LLMs to explain dataset&model. It's used by interp at OpenAI and societal impact Anthropic Tutorial here. It's a great direction & someone should carry the torch :) Thesis available, if you wanna read my acknowledgement section=P

New Tab model, 1M+ context windows, and a preview of our background agent

1) What

A conversation on the optimal reward for coding agents, infinite context models, and real-time RL

Nice conversation with Sualeh Asif Charlie Snell Aman Sanger Federico Cassano

Really enjoyed this conversation! A good look into how we're training frontier models at Cursor

new here

Cursor 1.0 is out now! Cursor can now review your code, remember its mistakes, and work on dozens of tasks in the background.

RL is very inference heavy and shifts infrastructure build outs heavily Scaling well engineered environments is difficult Reward hacking and non verifiable rewards are key areas of research Recursive self improvement already playing out Major shift in o4 and o5 RL training