AI300

Lilian Weng / Thinking Machines Cofounder
3661.
On-policy distillation provides an elegant way to use the teacher model as a process reward model to provide dense reward while preventing SFT style "OOD shock" during rollout.›
LLM score 90 · about 1 year ago
Mira Murati / Thinking Machines CEO
3662.
Combining the benefits of RL and SFT with on-policy distillation, a promising approach for training small models for domain performance and continual learning.›
LLM score 85 · about 1 year ago
Xiang Fu / Researcher at Periodic Labs
3663.
To adapt to today's world, crank down discount factor and crank up learning rate
LLM score 75 · about 1 year ago
John Schulman
3664.
Happy to share a new paper! Designing model behavior is hard -- desirable values often pull in opposite directions.›
LLM score 80 · about 1 year ago
Xiang Fu / Researcher at Periodic Labs
3665.
Every AI researcher should read two essays.›
LLM score 80 · about 1 year ago
Sally Zhu / Researcher at Flapping Airplanes
3666.
🔎Did someone steal your language model? We can tell you, as long as you shuffled your training data🔀.›
LLM score 95 · about 1 year ago
Aleksander Madry / OpenAI Researcher
3667.
Does removing a data source improve model capabilities? For instance, does adding another language data source increase model toxicity?›
LLM score 75 · about 1 year ago
Noam Brown / OpenAI Research Scientist
3668.
Below is a deep dive into why self play works for two-player zero-sum (2p0s) games like Go/Poker/Starcraft but is so much harder to use in "real world" domains.›
LLM score 80 · about 1 year ago
Noam Brown / OpenAI Research Scientist
3669.
@pli_cachete @karpathy Because it means self play converges to a minimax equilibrium which is essentially an unbeatable strategy.›
LLM score 80 · about 1 year ago
Noam Brown / OpenAI Research Scientist
3670.
Self play works so well in chess, go, and poker because those games are two-player zero-sum.›
LLM score 75 · about 1 year ago
Noam Brown / OpenAI Research Scientist
3671.
.@Stanford courses are high-quality but the policies are definitely outdated.›
LLM score 20 · about 1 year ago
Demis Hassabis / CEO of DeepMind
3672.
Awesome to see Veo 3.1 top the LMArena video leaderboards by a large distance with big improvements over Veo 3.0 for text-to-video (+30) and image-to-video (+70)! 🔥Huge congrats to the team! Try it for yourself in https://t.co/QgTpxTL8DQ and the @GeminiApp https://t.co/nQ14ZnkKnr
LLM score 20 · about 1 year ago
Noam Brown / OpenAI Research Scientist
3673.
@fchollet Didn't you say just two months ago that you think AGI is about 5 years away?›
LLM score 20 · about 1 year ago
Richard Song / DeepMind Research Scientist
3674.
Add-on: from our experience with Regression Language Models which absorb evaluations via weight updates - it takes at least *thousands* of (x,y) pairs to learn any meaningful world model of the objective.›
LLM score 92 · about 1 year ago
Demis Hassabis / CEO of DeepMind
3675.
Super excited to be collaborating with Commonwealth Fusion Systems @CFS_energy to use AI to accelerate fusion development - and move closer to a sustainable future with limitless clean energy https://t.co/kCS1UKEtjP
LLM score 70 · about 1 year ago
Demis Hassabis / CEO of DeepMind
3676.
Veo 3 is the state-of-the-art in video models.›
LLM score 30 · about 1 year ago
Demis Hassabis / CEO of DeepMind
3677.
We processed over 1.3 Quadrillion tokens last month - that's 1,300,000,000,000,000 tokens! or to put it another way that's 500M tokens a second or 1.8 Trillion tokens an hour...›
LLM score 75 · about 1 year ago
Geoffrey Hinton
3678.
Some generous companies in Toronto are funding three lectures on AI safety by Owain Evans on Nov 10, 11, 12.›
LLM score 60 · about 1 year ago
Tri Dao / Chief Scientist at Together
3679.
This work, led by @_junxiong_wang and @ben_athi, is a first step towards building AI systems that evolve and get better as you use them.›
LLM score 20 · about 1 year ago
Noam Brown / OpenAI Research Scientist
3680.
@itsandrewgao Worst-of-N is a good solution to this.›
LLM score 85 · about 1 year ago
Demis Hassabis / CEO of DeepMind
3681.
More info on some the cool capabilities of Genie 3 here: https://t.co/wcKztrUSu6
LLM score 40 · about 1 year ago
Demis Hassabis / CEO of DeepMind
3682.
Fantastic to see Genie 3, our state-of-the-art world model, featured in @TIME's 2025 Best Inventions.›
LLM score 40 · about 1 year ago
Hang Gao / ex MTS at xAI
3683.
Excited to share what I’ve been working on since joining @xai — Grok Imagine v0.9.›
LLM score 85 · about 1 year ago
Noam Brown / OpenAI Research Scientist
3684.
@littmath I decided to check with a doctor friend and they told me “NSAIDs have a tendency to increase your risk of clotting (despite the fact that they affect the functionality of your platelets resulting in a tendency to bleed)”
LLM score 40 · about 1 year ago
Noam Brown / OpenAI Research Scientist
3685.
@littmath Can you point out what specifically in the GPT-5 Thinking response you find problematic? Because so far you’ve not provided any evidence that its critique of “prevents blood clots” is unjustified
LLM score 20 · about 1 year ago
Noam Brown / OpenAI Research Scientist
3686.
@littmath I'm confused why you're so confidently taking a strong position on this despite admitting that you're not a doctor.›
LLM score 20 · about 1 year ago
Noam Brown / OpenAI Research Scientist
3687.
Are these just hallucinations by GPT-5 Thinking? Nope.›
LLM score 40 · about 1 year ago
Chi Jin / OpenAI Researcher
3688.
🚀With early access to Tinker, we matched full-parameter SFT performance as in Goedel-Prover V2 (32B) (on the same 20% data) using LoRA + 20% of the data.›
LLM score 92 · about 1 year ago
Noam Brown / OpenAI Research Scientist
3689.
"One of our goals is to discover superconductors that work at higher temperatures than today's materials"›
LLM score 70 · about 1 year ago
Jason Wei / AI Researcher at Meta
3690.
Bullish, in the coming decades majority of compute will be spent on ai for science https://t.co/8aLDhazs6i
LLM score 70 · about 1 year ago