art, paintings and painters
A neuronOne of thousands of tiny dials inside each layer. It reads the text so far and produces a number, its activation, that nudges what the model writes next. in layerOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. 6 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featuredA neuron picked for the full treatment: steering outputs and a written explanation. About 3,000 of the 86,000 neurons are featured., so it has steering outputs.
What this neuron does
This neuron fires on words about art, such as 'sampling', 'urban' in a painting context, and 'character' in illustration. It goes strongly negative on the names of painters like Monet and Van Gogh. Turned up, the model's writing kept repeating numbers and temperatures rather than art.
Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).
Fires most on
The text windows where this neuron's activationThe number a neuron produces for a token. Large positive means it fired hard; negative means it pushed the other way. was highest. Amber marks positive activation; the outlined tokenA piece of text the model reads at once: often a word, sometimes part of a word or a punctuation mark. is the peak. Hover a token for its strength.
-
2007. The result is encouraging. The model provides an objective basis for the development of acid rain forecasting operation in the province.⏎This is a preview of subscription content
-
is anchored in time and history, but it also provides a broad tour of music from the global movement against the brutal Apartheid regime. We’ve collected a sampling of songs from traditional South African folk songs, and protest songs from famous musicians from
-
or a favorite teacher.⏎- Think of famous people and take turns introducing them.⏎Copyright © 2004 by Susan Kettmann. Excerpted from The 2,000 Best Games & Activities with permission of its publisher
-
paintings were of peasant life in sombre colours. Once settled in Paris, however, his paintings took on new painting styles and brighter colours, and he started painting urban and suburban subjects. He liked the bright colours Impressionists used, and left
-
period (1603-1868). It is a form of woodblock printing that was popular among the common people of Japan, depicting scenes from everyday life, landscapes, and popular culture. Ukiyo-e prints were affordable
-
own characters first? You will get better faster if you! This article will go over some basic steps to get you started with digital character illustration.⏎Tips for digital character<|endoftext|>A city of ancient Sumer in southern Mesopotamia on a site
SteeringAdding a push along a neuron while the model writes, to see how its output changes. Turning it up or down.
Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.
Here is something I have been thinking about lately: the pieces are a mix of watercolor and oil paintings, and I am currently working on the final work. The painting is titled "The Last Days of a Painted Figure". I have been painting this work in my own words for the
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Charles Gleyre (1806 – 1874) (Gleyre Marc-Gabriel trained students such as Renoir, Monet, Sisley), he attended from 1875 training in the
-
- blue paper⏎- black paper⏎- pencil top eraser⏎The piece we selected to recreate was Van Gogh’s Starry Night. We’ve been learning about apples this week, an easy
-
to various mediums of art and famous artists and pieces of art. This week we did a quick lesson using this adorable book called Touch the Art: Make Van Gogh’s Bed by Julie Appel and Amy Guglielmo.⏎Touch the
-
steps go through blank cells or even part of an entirely different shape.⏎In a stroke of genius, Leo realised that the most important aspect of a post-cubist squarist painting is the largest diameter of any shape. He announces that there
Connections
Show on mapWhich neuronsOne of thousands of tiny dials inside each layer. It reads the text so far and produces a number, its activation, that nudges what the model writes next. in earlier layersOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.
Fed by earlier layers
- L4:807 suppresses
- L3:993 suppresses
- L4:1264 excites
- L4:1727 suppresses
- L5:198 excites unclear: weak "and" in lists
- L5:62 excites
- L4:2220 weakly suppresses
- L5:94 weakly suppresses
- L3:691 weakly excites
- L4:2541 weakly excites
- L3:122 weakly excites
- L2:1657 weakly excites
- L5:84 weakly suppresses
- L1:1464 weakly excites
- L5:2922 weakly excites
- L4:1252 weakly excites
- L5:1818 weakly suppresses
- L5:293 weakly excites
- L4:2680 weakly suppresses
- L1:2801 weakly excites
- L1:331 weakly suppresses
- L3:464 weakly excites
- L5:132 weakly suppresses
- L0:1 weakly suppresses
Feeds later layers
- L23:2973 suppresses
- L10:2348 suppresses
- L20:2649 suppresses
- L7:1801 suppresses
- L9:1925 suppresses
- L20:2050 suppresses
- L21:2123 suppresses
- L7:2298 suppresses
- L21:1229 suppresses
- L25:1584 suppresses art and artist names in lists
- L23:1523 suppresses
- L24:1561 suppresses
- L20:2537 suppresses
- L26:360 suppresses
- L24:1754 suppresses
- L20:1990 suppresses
- L26:3031 suppresses
- L8:2369 suppresses
- L20:319 excites
- L24:2646 suppresses
- L27:1279 suppresses
- L9:907 suppresses
- L10:505 suppresses
- L26:1739 suppresses
Activation histogramA chart of how often a neuron takes each activation value.
How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.
Logit lensReading which output tokens a neuron pushes toward or away from, straight off its output weights, before the later layers have their say.
The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layersOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. instead.
Promotes
- ellan
- ivia
- uada
- icana
- swick
- ickers
- Crunch
- Kimber
- igration
- ellers
Suppresses
- paints
- paintings
- 艺术品
- painting
- paint
- 艺术家
- artistic
- artists
- artworks
- 作品