Hierarchical Reasoning Model: Substance or Hype?
๐ Free resources (reading list + visuals): https://www.patreon.com/c/JuliaTurc
๐ HRM paper: https://arxiv.org/abs/2506.21734
โถ๏ธ Yacine's YouTube channel: https://www.youtube.com/@deeplearningexplained
In this video, we dive into the Hierarchical Reasoning Model (HRM), a new architecture from Sapient Intelligence that challenges scaling as the only way to advance AI. With only 27M parameters, 1000 training examples, and no pretraining, HRM still manages to place on the notoriously difficult ARC-AGI leaderboard, right next to models from OpenAI and Anthropic.
Together with Yacine Mahdid (neuโฆ
Watch on YouTube โ
(saves to browser)
Chapters (12)
Introducing HRM
1:23
Why Sudoku breaks Transformers
3:07
Recurrence via Chain-of-Thought
4:22
HRM: bird's eye view
6:30
Latent recurrence
8:23
The neuroscience backing
11:43
The H and L modules
12:32
Backprop-through-time approximation
13:48
The outer loop
19:31
Training data augmentation
22:59
Evaluation on Sudoku
24:07
Evaluation on ARC-AGI
DeepCamp AI