Hierarchical Reasoning Model
cs.AI, cs.LG
Submitted: 2025-06-26
Updated: 2026-09-30
Code: https://github.com/sapientinc/HRM
Project page: https://t-dillon.github.io/tdoku
License: http://creativecommons.org/licenses/by/4.0/
The gist: Reasoning, the process of devising and executing complex goal-oriented action sequences, remains a critical challenge in AI.
Terminology
Abstract
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.
Sources
- Transformers meet Neural Algorithmic Reasoners
- Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
- Premise Order Matters in Reasoning with Large Language Models
- Will we run out of data? Limits of LLM scaling based on human-generated data
- Insights into Transient Dynamics of Bacteria Laden Liquid Bridges
- On the Measure of Intelligence
- ARC Prize 2024: Technical Report
- ARC-AGI-2: A New Challenge for Frontier AI Reasoning Systems
- On Training Implicit Models
- SHINE: SHaring the INverse Estimate from the forward pass for bi-level optimization and implicit models
- Playing Atari with Deep Reinforcement Learning
- Implicit Bias of AdamW: $\ell_\infty$ Norm Constrained Optimization
- GLU Variants Improve Transformer
- Root Mean Square Layer Normalization
- Large Language Model Guided Tree-of-Thought
- Learning Iterative Reasoning through Energy Diffusion
- Sudoku-Bench: Evaluating creative reasoning with Sudoku variants
- Continuous Thought Machines
- Goat: Fine-tuned LLaMA Outperforms GPT-4 on Arithmetic Tasks
- Adaptive Computation Time for Recurrent Neural Networks
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Agents' Last Exam
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems