vdlad/Remarkable-Robustness-of-LLMs

Codebase the paper "The Remarkable Robustness of LLMs: Stages of Inference?"

29
/ 100
Experimental

This codebase allows AI researchers to conduct in-depth analyses of Large Language Models (LLMs) to understand their internal workings and robustness. It takes a pre-trained LLM and various experimental configurations as input, then outputs dataframes, visualizations, and metrics detailing neuron activity, attention patterns, and model behavior under interventions. Researchers and alignment engineers who study model interpretability and reliability would use this.

No commits in the last 6 months.

Use this if you need to perform detailed ablation studies, analyze neuron activations, or visualize attention mechanisms within TransformerLens-supported LLMs to understand their decision-making processes.

Not ideal if you are a practitioner solely focused on deploying or fine-tuning LLMs without needing to delve into their internal, mechanistic interpretability.

LLM interpretability AI safety research Mechanistic interpretability Model robustness Neural network analysis
Stale 6m No Package No Dependents
Maintenance 2 / 25
Adoption 6 / 25
Maturity 16 / 25
Community 5 / 25

How are scores calculated?

Stars

19

Forks

1

Language

Jupyter Notebook

License

MIT

Last pushed

Jun 11, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/vdlad/Remarkable-Robustness-of-LLMs"

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