The gap between open and closed models are narrowing and I expect this trend to continue. As foundation models become commoditized on a global level, the most interesting directions from both research and commercial is not in their development but in finding new ways to use them!
@hardmaru
-
Evolution Strategies at Scale Outperforms PPO for LLM Fine-Tuning
By
–
Evolution Strategies can be applied at scale to fine-tune LLMs, and outperforms PPO and GRPO in many model settings!
— hardmaru (@hardmaru) 7 octobre 2025
Fantastic paper “Evolution Strategies at Scale: LLM Fine-Tuning Beyond Reinforcement Learning” by @yule_gan, Risto Miikkulainen and team.https://t.co/CEyX6Z5ulG https://t.co/mPkXABJcQzEvolution Strategies can be applied at scale to fine-tune LLMs, and outperforms PPO and GRPO in many model settings! Fantastic paper “Evolution Strategies at Scale: LLM Fine-Tuning Beyond Reinforcement Learning” by @yule_gan
, Risto Miikkulainen and team. https://
arxiv.org/abs/2509.24372 -
Introducing Continuous Thought Machines New Framework
By
–
Introducing Continuous Thought Machineshttps://t.co/Fjwl82PZ9E
— hardmaru (@hardmaru) 6 octobre 2025
(Blog post from earlier this year.)
Summary tweet:https://t.co/bOGIDVDZodIntroducing Continuous Thought Machines https://
sakana.ai/ctm/
(Blog post from earlier this year.) Summary tweet: -
Continuous Thought Machines Spotlight at NeurIPS 2025
By
–
Proud to share that “Continuous Thought Machines” will be presented as a spotlight at #NeurIPS2025 ✨
— hardmaru (@hardmaru) 6 octobre 2025
Updated paper: https://t.co/ZWiidb1eTX
Interactive Demo: https://t.co/d6hqid60t7
GitHub: https://t.co/O0FS9jM8gk https://t.co/Zgv6TIFybSProud to share that “Continuous Thought Machines” will be presented as a spotlight at #NeurIPS2025 Updated paper: https://
arxiv.org/abs/2505.05522
Interactive Demo: https://
pub.sakana.ai/ctm/
GitHub: https://
github.com/SakanaAI/Conti
nuous-Thought-Machines
… -
ShinkaEvolve: Cost-Efficient AlphaEvolve Implementation and Comparison
By
–
ShinkaEvolve is our implementation of AlphaEvolve (we try to make it more sample efficient to save costs). Here is a comparison between Darwin Gödel Machine and AlphaEvolve from @richardcsuwandi
’s blog post discussing both methods: https://
richardcsuwandi.github.io/blog/2025/dgm/ -
ShinkaEvolve: Open-Source Framework for Scientific Program Evolution
By
–
Proud to release ShinkaEvolve, our open-source framework that evolves programs for scientific discovery with very good sample-efficiency! 🐙
— hardmaru (@hardmaru) 25 septembre 2025
Paper: https://t.co/05rjMiSxOL
Blog: https://t.co/yjuvV0xpif
Project: https://t.co/DpSEScrAws https://t.co/2pmiPesV4KProud to release ShinkaEvolve, our open-source framework that evolves programs for scientific discovery with very good sample-efficiency! Paper: https://
arxiv.org/abs/2509.19349
Blog: https://
sakana.ai/shinka-evolve/
Project: https://
github.com/SakanaAI/Shink
aEvolve
… -
Wayve autonomous vehicle successfully navigates Tokyo rush hour streets
By
–
It was great to see you around, @alexgkendall
! I’m so impressed that the @wayve_ai autonomous vehicle navigated flawlessly in crowded Tokyo streets during rush hour, especially inside the shotengai (商店街) streets with lots of pedestrian traffic. -
Adaptive Branching Tree Search for LLM Inference Scaling
By
–
“Wider or Deeper? Scaling LLM Inference-Time Compute with Adaptive Branching Tree Search” has been accepted as a spotlight at #NeurIPS2025 ✨
— hardmaru (@hardmaru) 22 septembre 2025
Paper: https://t.co/Xj6uLJITe4
Blog: https://t.co/3qSUaEixQU
GitHub: https://t.co/fo6LxAf9AP https://t.co/vIX0t5bbOn“Wider or Deeper? Scaling LLM Inference-Time Compute with Adaptive Branching Tree Search” has been accepted as a spotlight at #NeurIPS2025 Paper: https://
arxiv.org/abs/2503.04412
Blog: https://
sakana.ai/ab-mcts/
GitHub: https://
github.com/SakanaAI/TreeQ
uest
… -
DGM Self-Evolving on ARC-AGI Task
By
–
Nice work! Pretty cool to see DGM self-evolving on ARC-AGI. @jennyzhangzt @shengranhu @cong_ml @RobertTLange @jeffclune
