Sakana AI ORG

The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery 🧑‍🔬

A system that uses large language models to automate the full scientific research cycle, generating hypotheses, running experiments, and producing formatted academic papers with AI peer review.

The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search

Continuous Thought Machines, because thought takes time and reasoning is a process.

Official repository of Evolutionary Optimization of Model Merging Recipes

Hypernetworks that adapt LLMs for specific benchmark tasks using only textual task description as the input

ShinkaEvolve: Towards Open-Ended and Sample-Efficient Program Evolution 🧬

A Self-adaptation Framework🐙 that adapts LLMs for unseen tasks in real-time!

Hypernetworks that update LLMs to remember factual information

A Tree Search Library with Flexible API for LLM Inference-Time Scaling

Automating the Search for Artificial Life with Foundation Models!

Training teachers with reinforcement learning able to make LLMs learn how to reason for test time scaling.

Code to train and evaluate Neural Attention Memory Models to obtain universally-applicable memory systems for transformers.

No description.

Cuda kernels for leveraging LLM sparsity to improve throughput and decrease the memory requirements during inference and training.

DiffusionBlocks: Block-wise Neural Network Training via Diffusion Interpretation

Extending the Context of Pretrained LLMs by Dropping Their Positional Embedding

Digital Red Queen: Adversarial Program Evolution in Core War with LLMs

Code for Discovering Preference Optimization Algorithms with and for Large Language Models

The official repository of ALE-Bench

The code repository of the paper: Competition and Attraction Improve Model Fusion

Browser-based chat UI for TinySwallow-1.5B that runs without API calls.

Official implementation of "TAID: Temporally Adaptive Interpolated Distillation for Efficient Knowledge Transfer in Language Models"

No description.

No description.

No description.

Interactive multi-agent NCA ecosystem simulation

RePo: Language Models with Context Re-Positioning

No description.

Python-based chat demo for TinySwallow-1.5B that works completely offline

CycleQD is a framework for parameter space model merging.

LLMs for Constructed Languages

Reimagining Agent-based Modeling with Large Language Model Agents via Shachi

edinet2dataset is a tool to construct financial dataset using EDINET.

[ICLR 2026] Evaluating the performance of LLMs on Japanese challenging financial tasks.

Data and code for understanding and generation of Kamon.

A high-throughput and memory-efficient inference and serving engine for LLMs

No description.

Large language models to diffusion finetuning code

Reasoning-based Evaluation and Ranking of Translations.

Code for Fast-weight Product Key Memory (FwPKM)

Neuroevolution Community

No description.

Learning Neural Cellular Automata that produce Open-Ended Alife!

Replication of the MNIST experiments from paper **RL's Razor: Why Online Reinforcement Learning Forgets Less**

An AI benchmark for creative, human-like problem solving using Sudoku variants

No description.

No description.

Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞 Now with more orca.

A DeepSpeed-based framework for distributed training and inference of language models.

Benchmarking Agentic LLM and VLM Reasoning On Games

No description.

MLE-bench is a benchmark for measuring how well AI agents perform at machine learning engineering

No description.

KamonBench: A Grammar-Based Dataset for Evaluating Compositional Factor Recovery in Vision-Language Models