MaleCNS v1.0 · Drosophila CNS · computational study

FRUIT FLYBRAINvs AI

Can neural wiring evolved by biology provide useful computational structure for artificial intelligence?

NeuroWeave compares connectome-topology-constrained neural architectures derived from the MaleCNS v1.0 Drosophila connectome against conventional artificial neural architectures under controlled tasks.

Connectome source
MaleCNSv1.0
Prototype subgraph
150nodes
Confirmed benchmarks
2tasks
Bio, control, conventional
10architectures

Prototype subgraphs of 150 nodes are extracted from the connectome and used as a fixed connectivity mask. Every architecture is evaluated across multiple random seeds; reported spread is the standard deviation across those seeds.

Comparison families: connectome-derived architectures, topology controls that match the graph statistics without the biological wiring, and conventional artificial baselines. No claim is made about living animals, complete nervous systems, or consciousness.

Current headline result · T-002
94.8%± 6.2

A1-BIO with SA-010 temporal dynamics on the delayed-recall task. Conventional baselines evaluated here reach 47.4% (MLP) and 57.6% (LSTM).

Configuration selected via exploratory sensitivity analysis. Task-specific finding — not a universal biological conclusion.

The comparison
BIOLOGICAL CONNECTOMEMALECNS V1.0 · DROSOPHILA CNSMALECNS-DERIVED ARCHITECTURECONNECTOME-TOPOLOGY-CONSTRAINEDCONTROLLED BENCHMARKT-001 · T-002 · MULTI-SEEDBIOLOGICAL TOPOLOGYCONVENTIONAL AI MODELS