# Llama Scope: Extracting Millions of Features from Llama-3.1-8B with Sparse Autoencoders

Preprint · Zhengfu He, Wentao Shu, Xuyang Ge, Lingjie Chen, Junxuan Wang, Yunhua Zhou, Frances Liu, Qipeng Guo, Xuanjing Huang, Zuxuan Wu, Yu-Gang Jiang, Xipeng Qiu

- Archive id: arxiv:2410.20526
- Published: 2024-10-27
- arXiv: https://arxiv.org/abs/2410.20526
- Citations: 5
- Source URL: https://arxiv.org/abs/2410.20526
- PDF: https://arxiv.org/pdf/2410.20526
- Provenance: arxiv

## Abstract

Sparse Autoencoders (SAEs) have emerged as a powerful unsupervised method for extracting sparse representations from language models, yet scalable training remains a significant challenge. We introduce a suite of 256 SAEs, trained on each layer and sublayer of the Llama-3.1-8B-Base model, with 32K and 128K features. Modifications to a state-of-the-art SAE variant, Top-K SAEs, are evaluated across multiple dimensions. In particular, we assess the generalizability of SAEs trained on base models to longer contexts and fine-tuned models. Additionally, we analyze the geometry of learned SAE latents, confirming that \\emph{feature splitting} enables the discovery of new features. The Llama Scope SAE checkpoints are publicly available at~\\url{https://huggingface.co/fnlp/Llama-Scope}, alongside our scalable training, interpretation, and visualization tools at \\url{https://github.com/OpenMOSS/Language-Model-SAEs}. These contributions aim to advance the open-source Sparse Autoencoder ecosystem and support mechanistic interpretability research by reducing the need for redundant SAE training.

_Abstract reproduced under CC0-1.0 (arXiv descriptive metadata)._

Topics: [Mechanistic Interpretability](https://space-kitty.com/topics/interpretability)

Cited in dossiers:
- [Mechanistic Interpretability: State of the Field (October 2026)](https://space-kitty.com/dossiers/interpretability)

---
Space Kitty Archive · https://space-kitty.com · API: https://space-kitty.com/api/v1 · MCP: https://space-kitty.com/mcp · Premium: $0.20/request via x402 (Base Sepolia (testnet)). Metadata: arXiv (CC0), OpenAlex (CC0), Europe PMC (CC BY/CC0 works), ClinicalTrials.gov.