The benchmark continues to reveal gaps between AI computation and human-like reasoning.

🔗 Blogpost: pub.sakana.ai/sudoku-gpt5/
📊 Leaderboard: pub.sakana.ai/sudoku/
📄 Report: arxiv.org/abs/2505.16135
💻 GitHub: github.com/SakanaAI/Sudoku-Bench

Our GRPO and "Thought Cloning" experiments (learning from expert solvers) show current methods struggle with spatial reasoning and creative insights humans use naturally.

Unlike Chess or Go, these puzzles require understanding novel rules through meta-reasoning, then maintaining consistency over long reasoning chains.

Learn more about our approach.

GitHub: github.com/SakanaAI/pet...
Online Technical Report: pub.sakana.ai/pdnca

Petri Dish Neural Cellular Automata (PD-NCA) is a new ALife substrate that consists of a differentiable world where multiple NCA learn to self-replicate and grow via ongoing gradient descent. Every individual is constantly trying to grow, all the while learning to adapt and out-compete its neighbors

How the ‘Attention is all you need’ paper was born from freedom, not pressure:

There’s a fairly wide gulf in capabilities both among different LLMs and different linguistic specifications, with it being notably easier for systems to deal with settings that are commoner cross-linguistically than those that are rarer.

PDF arxiv.org/abs/2510.07591
Code github.com/SakanaAI/IASC

Our goals with IASC:

1/ We hope that these tools will be fun to use for creating artificially constructed languages.

2/ We are interested in exploring what LLMs ‘know’ about language—not what they know about any particular language, but how much they know about and understand linguistic concepts.

We are happy to announce the release of IASC, an Interactive Agentic System for ConLangs (Constructed Languages).

GitHub: github.com/SakanaAI/IASC

By making ShinkaEvolve open-source, our goal is to democratize access to advanced discovery tools. We envision it as a companion to help scientists and engineers, building efficient, nature-inspired systems to unlock the future of AI research.

GitHub Project: github.com/SakanaAI/Shi...

ShinkaEvolve's efficiency comes from three key innovations:

1) Adaptive parent sampling to balance exploration and exploitation.

2) Novelty-based rejection filtering to avoid redundant work.

3) A bandit-based LLM ensemble that dynamically picks the best model for the job.

3/ LLM Training: It discovered a novel load balancing loss for MoE models, improving model performance and perplexity.

2/ Competitive Programming: On ALE-Bench, it improved an existing agent's solution, turning a 5th place result into a 2nd place leaderboard rank for one task.

We applied ShinkaEvolve to a diverse set of hard problems:

1/ AIME Math Reasoning: It evolved sophisticated agentic scaffolds that significantly outperform strong baselines, discovering a Pareto frontier of solutions trading performance for efficiency.

On the classic circle packing optimization problem, ShinkaEvolve discovered a new state-of-the-art solution using only 150 samples. This is a massive leap in efficiency compared to previous methods that required thousands of evaluations.

Many evolutionary AI systems are powerful but act like brute-force engines, burning thousands of samples to find good solutions. This makes discovery slow and expensive. We took inspiration from the efficiency of nature. ‘Shinka’ (進化) is Japanese for evolution.