Hierarchical Reasoning Model

Reasoning, the process of devising and executing complex goal-oriented action sequences, remains a critical challenge in AI. Current large language models (LLMs) primarily employ Chain-of-Thought (CoT) techniques, which suffer from brittle task decomposition, extensive data requirements, and high latency. Inspired by the hierarchical and multi-timescale processing in the human brain, we propose the Hierarchical Reasoning Model (HRM), a novel recurrent architecture that attains significant computational depth while maintaining both training stability and efficiency. HRM executes sequential reasoning tasks in a single forward pass without explicit supervision of the intermediate process, through two interdependent recurrent modules: a high-level module responsible for slow, abstract planning, and a low-level module handling rapid, detailed computations. With only 27 million parameters, HRM achieves exceptional performance on complex reasoning tasks using only 1000 training samples. The model operates without pre-training or CoT data, yet achieves nearly perfect performance on challenging tasks including complex Sudoku puzzles and optimal path finding in large mazes. Furthermore, HRM outperforms much larger models with significantly longer context windows on the Abstraction and Reasoning Corpus (ARC), a key benchmark for measuring artificial general intelligence capabilities. These results underscore HRM’s potential as a transformative advancement toward universal computation and general-purpose reasoning systems.

Read Our Paper: https://arxiv.org/abs/2506.21734

Join Our Discord Community: https://discord.gg/sapient

Quick Start Guide 🚀

Prerequisites ⚙️

Ensure PyTorch and CUDA are installed. The repo needs CUDA extensions to be built. If not present, run the following commands:

# Install CUDA 12.6
CUDA_URL=https://developer.download.nvidia.com/compute/cuda/12.6.3/local_installers/cuda_12.6.3_560.35.05_linux.run

wget -q --show-progress --progress=bar:force:noscroll -O cuda_installer.run $CUDA_URL
sudo sh cuda_installer.run --silent --toolkit --override

export CUDA_HOME=/usr/local/cuda-12.6

# Install PyTorch with CUDA 12.6
PYTORCH_INDEX_URL=https://download.pytorch.org/whl/cu126

pip3 install torch torchvision torchaudio --index-url $PYTORCH_INDEX_URL

# Additional packages for building extensions
pip3 install packaging ninja wheel setuptools setuptools-scm

Then install FlashAttention. For Hopper GPUs, install FlashAttention 3

git clone git@github.com:Dao-AILab/flash-attention.git
cd flash-attention/hopper
python setup.py install

For Ampere or earlier GPUs, install FlashAttention 2

pip3 install flash-attn

Install Python Dependencies 🐍

pip install -r requirements.txt

W&B Integration 📈

This project uses Weights & Biases for experiment tracking and metric visualization. Ensure you're logged in:

wandb login

Run Experiments

Quick Demo: Sudoku Solver 💻🗲

Train a master-level Sudoku AI capable of solving extremely difficult puzzles on a modern laptop GPU. 🧩

# Download and build Sudoku dataset
python dataset/build_sudoku_dataset.py --output-dir data/sudoku-extreme-1k-aug-1000  --subsample-size 1000 --num-aug 1000

# Start training (single GPU, smaller batch size)
OMP_NUM_THREADS=8 python pretrain.py data_path=data/sudoku-extreme-1k-aug-1000 epochs=20000 eval_interval=2000 global_batch_size=384 lr=7e-5 puzzle_emb_lr=7e-5 weight_decay=1.0 puzzle_emb_weight_decay=1.0

Runtime: ~10 hours on a RTX 4070 laptop GPU

Trained Checkpoints 🚧

• ARC-AGI-2
• Sudoku 9x9 Extreme (1000 examples)
• [Maze 30x30 Hard (1000 examples)](https://huggingface.co/sapientinc/HRM-checkpoint-maze-30