spiral and screw words
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. 2 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. 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 responds to words about spirals and screws, such as 'spiral', 'lead screw' and 'spiral fracture'. Turning it up or down did not change the writing in a clear, related way.
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.
-
pipetting systems.⏎One of the most common applications of lead screws is for positioning the print head in 3D printers. They often use 8mm lead screws as demonstrated in the video below. The concepts explained above are all shown in the
-
alcohol, bi-carbonate of soda, vaseline, or burnt with a match in an attempt to remove. Such action can induce a tick to release spirochetes and/or other pathogens known to cause
-
terms used to identify the types of broken bones, including:⏎TransverseA bone break is considered transverse when the fracture extends across the bone.⏎SpiralAs you may have guessed, a spiral fracture spirals around the bone.⏎
-
es, CNC machines or complex robotics, but they can just as commonly be found around the home, in conventional products such as printers or DVD drives.⏎A Lead Screw’s main purpose is to rotate, using its teeth (or single tooth) to
-
your home Child Friendly!⏎For more information about how you can reduce risks in your home and what to do if your child has an accident, visit⏎MySpira is the world's first augmented reality asthma training game, developed in collaboration with
-
the two sums are in general different, and the above formula does not necessarily hold.⏎It was shown by Lehmer that the exponential Gauss walk always has the spiral-type shape observed in the visualization. On the other hand, the shape of the
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: I am very much like the first of the first of the first of the first of the first of the first of the first of the first of the first of the first of the first of the first of the first of the first of the
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
-Caulle however a larger VEI3 eruption occurred in 1960. This has been attributed in various sources to the Valdivia earthquake of 1960 the most powerful earthquake ever recorded on the
-
of the Leopold Leadership Program at Stanford University. He is also the director of the Tree Biosafety
-
: Stepped stage⏎Learn about this topic in these articles:⏎invention by Jessner⏎...be attributed to the Expressionist movement. The director Leopold Jessner capitalized on the earlier innovations in stage design. His use of
-
ademia del Cimento in 1657. In the 18th century the Lorraine family kept collecting and the Grand Duke Peter Leopold of Habsburg-Lorraine founded in Torrigiani Palace a veritable Museum
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
- L1:951 excites
- L0:1224 suppresses
- L0:1301 weakly suppresses
- L0:178 weakly suppresses
- L1:1960 weakly excites
- L1:2657 weakly excites
- L1:189 weakly suppresses
- L1:1460 weakly excites
- L0:2004 weakly suppresses
- L1:1965 weakly suppresses
- L0:653 weakly suppresses
- L1:2946 weakly suppresses
- L1:2728 weakly suppresses
- L1:1692 weakly excites
- L1:2143 weakly suppresses
- L1:545 weakly suppresses
- L1:2816 weakly excites
- L0:2942 weakly suppresses
- L1:1610 weakly excites
- L1:2902 weakly suppresses
- L1:492 weakly excites
- L1:1857 weakly suppresses
- L1:2164 weakly excites
- L1:5 weakly suppresses
Feeds later layers
- L9:1988 weakly suppresses
- L12:908 weakly excites
- L11:1319 weakly excites
- L7:502 weakly suppresses unclear: words after 'less than' or 'of'
- L14:3062 weakly excites
- L20:663 weakly excites
- L19:1081 weakly excites
- L9:1431 weakly excites
- L4:2181 weakly suppresses
- L4:1148 weakly suppresses
- L20:1092 weakly suppresses
- L17:1917 weakly excites
- L7:2471 weakly suppresses
- L23:2421 weakly excites
- L12:678 weakly excites
- L10:1436 weakly excites
- L3:1222 weakly excites
- L11:2493 weakly suppresses
- L23:1865 weakly suppresses
- L5:2381 weakly excites
- L7:341 weakly excites 'to' in 'according to'
- L8:2928 weakly excites
- L20:1644 weakly suppresses
- L13:2671 weakly 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
- [value
- [count
- ments
- //{⏎
- itchens
- clud
- cts
- //{⏎
- [vi
- رب
Suppresses
- STRUCTOR
- ệt
- privileged
- kB
- okit
- ham
- nore
- puck
- actory
- 杉