GitHub - pathwaycom/bdh: BDH (Dragon Hatchling) – Architecture and Code · GitHub
/" data-turbo-transient="true" />
Skip to content
Type / to search
Sign in<br>Sign upAppearance settings
You signed in with another tab or window. Reload to refresh your session.<br>You signed out in another tab or window. Reload to refresh your session.<br>You switched accounts on another tab or window. Reload to refresh your session.
Dismiss alert
{{ message }}
Uh oh!
There was an error while loading. Please reload this page.
pathwaycom
bdh
Public
Notifications<br>You must be signed in to change notification settings
Fork<br>233
Star<br>3.5k
main
BranchesTags
Go to file
CodeOpen more actions menu
Folders and files<br>NameNameLast commit message<br>Last commit date<br>Latest commit
History<br>28 Commits<br>28 Commits
figs
figs
.gitignore
.gitignore
LICENSE.md
LICENSE.md
README.md
README.md
bdh.py
bdh.py
requirements.txt
requirements.txt
train.py
train.py
View all files
Repository files navigation
BDH (Dragon Hatchling)
Bridging the Gap Between Transformers and the Brain
BDH (Dragon Hatchling) is a biologically inspired large language model architecture that connects principles of deep learning with the foundations of neuroscience. Developed by researchers at Pathway, BDH provides a theoretical and practical framework for understanding the emergence of reasoning and generalization in artificial systems.
This repository contains the official implementation from the paper:
A. Kosowski, P. Uznański, J. Chorowski, Z. Stamirowska, M. Bartoszkiewicz.<br>The Dragon Hatchling: The Missing Link between the Transformer and Models of the Brain, arXiv (2025).
Overview
BDH represents a scale-free, locally interacting network of neurons capable of intrinsic reasoning dynamics. BDH scales like a Transformer on performance benchmarks—yet retains full interpretability and theoretical grounding in the fine-grained dynamics of neuron interactions.
Key properties:
Scale-free network topology mimicking biological connectivity
Locally interacting neuron particles with excitatory/inhibitory dynamics
Hebbian working memory based on synaptic plasticity, displaying monosemanticity
GPU-friendly state-space formulation for efficient implementation
Interpretable activations that are sparse and positive
BDH formalizes a bridge between neural computation and machine-based language understanding . It shows how macro reasoning behavior in large AI models emerges from micro-level neuron dynamics , guided by principles of graph theory and local computation.
Empirically, BDH matches GPT-2–scale Transformers across language and translation tasks at equivalent parameter scales (10M–1B).
Architecture
Relation to Transformers
BDH and the Transformer share attention-inspired computation; however, BDH’s graph-based architecture makes its attention emerge naturally from neuron-level interactions , reflecting attention as seen in biological systems.
Scaling Laws
BDH follows Transformer-like scaling laws , maintaining parameter efficiency while achieving interpretability at any scale.
Latest research update: Sudoku Benchmark
Note: The Sudoku Extreme result refers to Pathway’s internal BDH implementation, not to the current open-source repository. This repository contains the implementation of the baseline variant as described in our public paper and does not reproduce the 97.4% benchmark result out of the box. See the dedicated Extreme Sudoku research blog post for additional benchmark context and the reported results.
On Sudoku Extreme, BDH reaches 97.4% accuracy across roughly 250,000 difficult puzzles, without chain-of-thought, solution backtracking, or external tool use, while leading LLMs struggle to perform on the benchmark at all.
Language is not enough for intelligence. Transformers process information token by token with limited internal state, which makes search-heavy, non-linguistic reasoning tasks like Sudoku awkward. BDH uses a larger latent reasoning space with intrinsic memory that supports learning and adaptation during use.
We believe that the future of AI will belong to systems that can reason natively across domains, that can hold multiple possibilities in a rich latent space, and that can converge on solutions without needing to verbalize every step. BDH is our answer to that challenge. It is designed to be a universal reasoning system that can speak our language without being trapped inside it. And yes, it solves Sudoku.
Read more: Post-transformers: Sudoku Bench
Performance Comparison
Model<br>Sudoku Extreme Accuracy<br>Relative Cost
Pathway BDH<br>97.4%<br>10× lower, No chain-of-thought
Leading LLMs (O3-mini, DeepSeek R1, Claude 3.7 8K)<br>~0%<br>High (chain-of-thought)
Table 1: Performance comparison on extreme Sudoku benchmarks (~250,000 difficult puzzles).
Source: Pathway internal data and https://arxiv.org/pdf/2506.21734 for the Leading LLMs’ accuracy score. Pathway’s approach reflects top-1 accuracy and does not rely on...