The Dragon Hatchling: The Missing Link between the Transformer and Models of the Brain
Adrian Kosowski, Przemysław Uznański, Jan Chorowski, Zuzanna Stamirowska, Michał Bartoszkiewicz
TL;DR
This work introduces Dragon Hatchling (BDH), a Large Language Model architecture grounded in scale-free, brain-inspired local graph dynamics. By modeling inference as edge-reweighting on neuron graphs and implementing a GPU-friendly BDH-GPU variant with ReLU-lowrank blocks and linear attention, the approach preserves Transformer-like scaling while improving interpretability and brain- plausibility. The authors establish theoretical links between BDH and brain models, show emergent modular, scale-free network structure, and demonstrate Transformer-like performance on language and translation tasks with 10M–1B parameter scales. They further demonstrate interpretability through monosemantic synapses, sparse activations, and a micro-foundational perspective on attention, arguing for axiomatic AI and potential brain-science insights. Overall, BDH offers a principled, scalable path to long-context reasoning with interpretable dynamics and practical GPU efficiency, bridging Transformer machinery and brain-inspired computation.
Abstract
The relationship between computing systems and the brain has served as motivation for pioneering theoreticians since John von Neumann and Alan Turing. Uniform, scale-free biological networks, such as the brain, have powerful properties, including generalizing over time, which is the main barrier for Machine Learning on the path to Universal Reasoning Models. We introduce `Dragon Hatchling' (BDH), a new Large Language Model architecture based on a scale-free biologically inspired network of \$n\$ locally-interacting neuron particles. BDH couples strong theoretical foundations and inherent interpretability without sacrificing Transformer-like performance. BDH is a practical, performant state-of-the-art attention-based state space sequence learning architecture. In addition to being a graph model, BDH admits a GPU-friendly formulation. It exhibits Transformer-like scaling laws: empirically BDH rivals GPT2 performance on language and translation tasks, at the same number of parameters (10M to 1B), for the same training data. BDH can be represented as a brain model. The working memory of BDH during inference entirely relies on synaptic plasticity with Hebbian learning using spiking neurons. We confirm empirically that specific, individual synapses strengthen connection whenever BDH hears or reasons about a specific concept while processing language inputs. The neuron interaction network of BDH is a graph of high modularity with heavy-tailed degree distribution. The BDH model is biologically plausible, explaining one possible mechanism which human neurons could use to achieve speech. BDH is designed for interpretability. Activation vectors of BDH are sparse and positive. We demonstrate monosemanticity in BDH on language tasks. Interpretability of state, which goes beyond interpretability of neurons and model parameters, is an inherent feature of the BDH architecture.