rhubarbwu/linguistic-collapse

Codebase for Linguistic Collapse: Neural Collapse in (Large) Language Models [NeurIPS 2024] [arXiv:2405.17767]

28
/ 100
Experimental

This project provides the tools and scripts to train and evaluate large language models (LLMs) like GPT-Neo, and then deeply analyze their internal representations for a phenomenon called 'Neural Collapse.' You can input your LLM configurations and training data, receive trained models, and then get detailed analysis results, including various metrics of linguistic collapse and visualization notebooks. It's designed for machine learning researchers and academics studying the fundamental behaviors of LLMs.

No commits in the last 6 months.

Use this if you are a machine learning researcher who wants to rigorously study the internal mechanics of how large language models learn and represent information.

Not ideal if you are looking for an off-the-shelf solution to train a production-ready LLM or to apply LLMs for practical tasks without deep architectural analysis.

machine-learning-research natural-language-processing neural-networks computational-linguistics model-analysis
No License Stale 6m No Package No Dependents
Maintenance 2 / 25
Adoption 6 / 25
Maturity 8 / 25
Community 12 / 25

How are scores calculated?

Stars

18

Forks

3

Language

Python

License

Last pushed

Apr 14, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/transformers/rhubarbwu/linguistic-collapse"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.