VISTA: A Panoramic View of Neural Representations
Tom White · 2024
Read VISTA: A Panoramic View of Neural Representations by Tom White (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-02412.txt.
Description
Tom White presents a 2024 tool for viewing what neural networks learn inside. Modern models store knowledge in vast multi-dimensional spaces. Reading those spaces directly is beyond human intuition. White's pipeline maps internal states into a two-dimensional semantic layout.
The pipeline, VISTA, builds collages from mapped representations. Patterns and links among internal features become visible at a glance. A case study applied the tool to sparse autoencoder features. Fresh traits and readings of those features emerged from the views. The paper reviews the method step by step. A public case study site accompanies the work. Implications span many areas of machine learning practice. The author frames visual exploration as a complement to numeric probes. Interpretability, he argues, needs tools the eye can use.
Catalog record
| CREATOR | Tom White |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 15 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-02412