skills.rest
choice5346
choice5346COMMUNITY
Explore all skills by this author →
GitHub

sparse-autoencoder-training

Trains and analyzes sparse-coded neural network features using SAELens and TransformerLens workflows.

Updated Aug 27, 2026
Search Tags:#feature discovery#language models#mechanistic interpretability#neural network analysis#saelens#sparse autoencoder
One-click install
npx skills add https://github.com/choice5346/BiSHE --skill sparse-autoencoder-training-choice5346
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: sparse-autoencoder-training
Source: https://github.com/choice5346/BiSHE/tree/main/.github/skills/saelens
Command: npx skills add https://github.com/choice5346/BiSHE --skill sparse-autoencoder-training-choice5346

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill provides tools and guidance for training and analyzing Sparse Autoencoders (SAEs) to decompose complex neural network activations into understandable, sparse features.

Core Features & Use Cases

  • Feature Discovery: Identify interpretable concepts learned by language models.
  • Mechanistic Interpretability: Analyze how models represent information, study superposition, and understand feature geometry.
  • SAE Training & Analysis: Offers workflows for both training custom SAEs and analyzing pre-trained ones.
  • Use Case: Discover what specific features in GPT-2 activations correspond to concepts like "sentiment," "grammar," or "named entities" to better understand model behavior.

Quick Start

Use the saelens skill to load a pre-trained SAE for GPT-2 small and analyze its features on a given text.

Frequently Asked Questions about sparse-autoencoder-training

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I train a sparse autoencoder for mechanistic interpretability research?▼

Sparse autoencoders decompose complex neural network activations into sparse, interpretable features, allowing you to identify concepts like sentiment or grammar learned by language models for mechanistic interpretability.

How does sparse autoencoder feature discovery work on language models?▼

Sparse autoencoder feature discovery works by decomposing language model activations into sparse representations, enabling you to identify specific interpretable concepts like named entities and analyze model behavior.

Can I analyze pre-trained sparse autoencoders for GPT-2 small activations?▼

Yes, you can analyze pre-trained sparse autoencoders for GPT-2 small activations by loading existing models with SAELens to evaluate and understand how the model represents information on given text.

Does SAELens support analyzing superposition and feature geometry in neural networks?▼

Yes, SAELens supports analyzing superposition and feature geometry in neural networks by facilitating the decomposition of activations into sparse, interpretable features to study how models represent information.

What is the best way to steer language model behavior based on learned sparse features?▼

The best way to steer language model behavior is by training sparse autoencoders to learn interpretable representations, enabling you to steer the model based on the discovered sparse features.

Related Skills

Explore More Software Engineering Skills→
jackspacejackspace
16

torch-geometric

Develop and train graph neural networks with PyTorch Geometric.

Community
Advanced
davpatel605-beepdavpatel605-beep

sparse-autoencoder-training

Train sparse autoencoders to decompose transformer activations into interpretable features.

Community
Advanced
GregsGreyCodeGregsGreyCode
6

sparse-autoencoder-training

Train and analyze sparse autoencoders on transformer activations using SAELens.

Community
Advanced
ihatesea69ihatesea69
3

sparse-autoencoder-training

Train and analyze Sparse Autoencoders on neural network activations using SAELens, TransformerLens, and PyTorch.

Community
Advanced