// LLMs Improving LLMs //
Interesting progress the past of couple of weeks around self-improving AI agents.
If autoresearch was interesting, you will like this read.
(bookmark it)
We've been hand-tuning test-time scaling for a year. This work asks what happens when you let an LLM search the space instead.
The paper introduces AutoTTS, a framework that reframes the human role: instead of designing branching, pruning, and stopping heuristics directly, you construct a discovery environment where TTS strategies can be searched automatically. They formulate width–depth TTS as controller synthesis over pre-collected reasoning trajectories and probe signals, so candidate controllers can be evaluated cheaply without repeated LLM calls.
Two design choices carry the search. Beta parameterization makes the control space tractable. Fine-grained execution-trace feedback tells the explorer LLM why a candidate failed, not just that it did.
On math reasoning benchmarks, the discovered controllers beat strong hand-designed baselines on the accuracy–cost Pareto frontier and generalize zero-shot to held-out benchmarks and model scales.
Entire discovery cost: $39.9 and 160 minutes.
Why it matters:
The era of researchers hand-crafting CoT, best-of-N, and self-consistency recipes is on a clock. Once the search loop is cheap enough, TTS becomes another thing LLMs do for themselves.
Paper: arxiv.org/abs/2605.08083
Learn to build effective AI agents in our academy: academy.dair.ai
134K Followers 333 FollowingAI research @AnthropicAI. Previously OpenAI & DeepMind.
Optimizing for a post-AGI future where humanity flourishes.
Opinions aren't my employer's.
3K Followers 1K FollowingAssistant Professor @imperialcollege
Working on Alignment & Safe AI
Ex: Principal Research Fellow @ucl, PhD @UniofOxford, Intern @Google, MEng @Polytechnique
13K Followers 519 FollowingCTO of Technology & Society @Google, working on fundamental AI research and exploring the nature and origins of intelligence. Order "What Is Intelligence?" ⬇️
53K Followers 232 FollowingYes, I *am* that ESR. Well, it's the question people usually ask.
Programmer, wandering philosopher, accidental anthropologist, troublemaker for liberty.
11K Followers 61 FollowingEmbedded Engineering @Google | Building https://t.co/7dhg3vg69m in spare time, enabling engineers reason well. From AI to Transistor. views are my own.
1.2M Followers 787 FollowingProfessor at NYU & Executive Chairman at AMI Labs.
Ex-Chief AI Scientist at Meta.
Researcher in AI, Machine Learning, Robotics, etc.
ACM Turing Award Laureate.
207K Followers 0 FollowingOG of: P and T in ChatGPT, 100x deeper learning, meta learning and RSI, neural distillation, GAN/World Model... Co-authored most-cited AI paper of 20th century
1.5M Followers 176 FollowingNobel Laureate. Co-Founder & CEO @GoogleDeepMind - working on AGI. Solving disease @IsomorphicLabs. Trying to understand the fundamental nature of reality.
270K Followers 185 FollowingCo-founder of Thinking Machines Lab @thinkymachines; Ex-VP, AI Safety & robotics, applied research @OpenAI; Author of Lil'Log
10K Followers 951 FollowingGeneral Partner @a16z building American Dynamism, formerly @PalantirTech. Up for any adventure but Arizona will always be home.
68K Followers 5K FollowingTechno-industrialist, believer in hard power and Pax Americana. Automated Factories and the New American Workforce are how we win.