neelsomani/symbolic-circuit-distillation
Automatically extract executable programs from pruned mechanistic circuits, extending OpenAI's Sparse Circuits
This project helps AI researchers understand the internal workings of small, specialized AI models by translating their complex 'neural circuits' into simple, human-readable computer programs. It takes a pruned circuit (like a diagram of interconnected neurons) as input and outputs a verified, executable algorithm. This tool is for scientists in mechanistic interpretability who are trying to pinpoint how AI models perform specific tasks.
Use this if you are an AI interpretability researcher trying to automatically translate a small, focused neural circuit from a transformer model into a clear, verifiable algorithm.
Not ideal if you are trying to interpret the full-scale behavior of a large language model or if you are not working with isolated, algorithmic circuits.
Stars
64
Forks
3
Language
Python
License
Apache-2.0
Category
Last pushed
Nov 23, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/neelsomani/symbolic-circuit-distillation"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
pykt-team/pykt-toolkit
pyKT: A Python Library to Benchmark Deep Learning based Knowledge Tracing Models
microsoft/archai
Accelerate your Neural Architecture Search (NAS) through fast, reproducible and modular research.
google-research/morph-net
Fast & Simple Resource-Constrained Learning of Deep Network Structure
AI-team-UoA/pyJedAI
An open-source library that leverages Python’s data science ecosystem to build powerful...
IDEALLab/EngiBench
Benchmarks for automated engineering design