Show HN: Fly Language Model
Category: ai-ml
Tags: language-model, connectome, recurrent-neural-networks, interpretability, computational-neuroscience
Score: 6.3/10 (Innovation: 7, Technical: 7, Documentation: 6, Utility: 5)
FLM (Fly Language Model) is an experimental language model trained from scratch on a connectome-derived recurrent network, using the fruit-fly brain's wiring as a fixed recurrent structure. It explores whether biological connectome topology and recurrent dynamics can support language modeling, with inspectable inference and controlled experiments comparing against GRU and transformer baselines. The project is interesting for its unique combination of neuroscience-inspired architecture and interpretable language modeling.
Target audience: AI researchers, computational neuroscientists, and machine learning engineers interested in biologically inspired architectures and interpretability
Repository: https://flm.kuber.studio/ · Python · 177 stars
View on Hacker News