London

Neel Nanda

Elite
@neelnanda-io

Mechanistic interpretability researcher at Google DeepMind. Formerly at Anthropic, formerly independent

1L-Sparse-Autoencoder. Python

139

Crosscoders. Python

59

Grokking. A Mechanistic Interpretability Analysis of Grokking

28

Neuroscope. Accompanying codebase for neuroscope.io, a website for displaying max activating dataset examples for language model neurons

14

neel-plotly. A very hacky set of functions for getting plotly to do what I want when doing mech interp research, designed to be compatible with PyTorch

14

Stable-Diffusion-Interp. Python

9

Synthetic-Word-Index-Embeddings. Python

8

Model-Diffing-Exploration. Python

7

LLaMA-Chat-Finetuning-Circuits. Python

6

mats_sae_training. Training Sparse Autoencoders on Language Models

6

Model-Family-Diffing. Python

5

Prediction-Suppression-Neurons.

5

SAE-Feature-Splitting. Python

4

Exploring-2L-SAE. Python

4

Tiny-Stories-SAEs. Python

4

GPT-Spelling-Bee. Python

4

CoT-Interp. Python

4

solu_project. Jupyter Notebook

3

Intro-to-MI-Techniques.

3

Qwen-MLP-Exploration. Python

3

tracr. Python

3

neurips-2025-site. HTML

3

SAE-Addition-Gemma. Python

3

MoverHeads. Mover Heads

3

CoT_Faithfulness_demo. Python

3

measuring-performance-of-OR-models. Comprehensive benchmarking tool for OpenRouter language models - measures performance across 70+ diverse tasks including text, code, math, and vision with detailed analysis

3

plotly_tensor. A wrapper around Plotly for various utilities with Mechanistic Interpretability Research

2

neelutils. Random utils for personal use

2

Testing-Gemma-2.

2

Retokenization-Probing. Python

2

Thinking-Diffing. HTML

2

Refusal-Replication. Python

2

Act-Add. Python

2

Finding-An-SAE-Circuit. Python

2

productivity_guard. A Gemini-powered productivity monitor that helps keep you focused by analyzing screenshots for procrastination

2

SoLU-Notebook-Test. Test for nbdev

2

Neuron-Families. Python

1

Coloured-Object-Subspaces. Python

1

Greater-Than. Python

1

S-Inhibition-Heads. Python

1

Object-Size-Comparison. Python

1

Variable-Binding.

1

neelnanda-io.github.io. HTML

1

TLWebsiteRedirect. HTML

1

Understanding-Arithmetic. Python

1

Copy-Suppression-Exploration. Python

1

An-Neuron. Python

1

Trigram-1L-Neurons. Python

1

ProbingRGB. Python

1
49
Apply