unclear: table pipes and markup symbols
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. 10 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
It mostly reacts to the vertical bars and slashes used to draw tables and diagrams, plus a few list-like words. The examples have little in common beyond that messy formatting. Turning it up gave repetitive, unrelated text.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
, then the node itself, and then its right child.⏎T / \ I S / \ D E<|endoftext|>Original URL: http://www.theregister.co.uk/2
-
5 percent⏎|Region||December 2013||December 2012||Percent Change|⏎|Region||December 2013||November 2013||Percent Change|<|endoftext|>In this post, we
-
Feast of Sacrifice|⏎|Saturday||September 02||Eid ul-Azha Day 2||Feast of Sacrifice|⏎|Saturday||September 30||Ashura||10th day of Muharr
-
the same.⏎- To this day we follow the model that Galileo created.⏎ag = g = 9.81m/s2⏎ag = g = acceleration due to gravity⏎Since gravity is just an acceleration like any other, it
-
lookup comes up quite often. In this article we explore 4 methods of how to conduct 2D lookups in Excel.⏎+++NEW+++NEW+++NEW+++ 2D XLOOKUP +++NEW+++NEW+++NEW+++⏎In a
-
юсь(⏎Y. Раскройте скобки, употребляя глаголы в Present, Past, Future Indefinite, Present, Past, Future Continuous. 1.When I (to open) the window, it
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 a student and I am going to write a report to my teacher, and I hope that you can figure out where I am. to find out the result of the first time, I will use a [your name] to write
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
"The Struggle for Power in the Ussr" StudyMode.com. 05 2012. 05 2012 <http://www.studymode.com
-
Struggle-For-Power-In-The-994323.html>.⏎"The Struggle for Power in the Ussr." StudyMode.com. 0<|endoftext|>Afghanistan and Colombia, half a world apart and different in culture
-
of Natural History – Tourist Attractions in New York City⏎American Museum of Natural History⏎American Museum of Natural History⏎The American Museum of Natural History is a landmark on the Upper West Side, Manhattan
-
whole current is then transferred to the trip coil which operates to trip the circuit breaker.⏎Email Based, Online Homework Assignment Help in Classification Of Relays⏎Transtutors is the best place to get answers to all your doubts regarding classification of relays
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
- L3:238 weakly suppresses
- L8:13 weakly suppresses
- L9:1658 weakly excites
- L7:349 weakly excites
- L8:7 weakly excites
- L4:234 weakly excites
- L8:3 weakly excites
- L0:892 weakly excites
- L7:49 weakly suppresses
- L9:911 weakly excites
- L2:1685 weakly suppresses
- L5:216 weakly suppresses
- L4:3018 weakly excites
- L3:202 weakly suppresses
- L8:3061 weakly suppresses
- L5:1076 weakly suppresses
- L2:48 weakly suppresses
- L5:1914 weakly suppresses
- L1:82 weakly suppresses
- L0:2971 weakly suppresses
- L5:80 weakly excites
- L4:2159 weakly excites
- L6:1063 weakly excites
- L8:1489 weakly suppresses
Feeds later layers
- L17:2956 suppresses
- L12:796 weakly excites
- L24:763 weakly excites
- L16:1081 weakly excites
- L14:1741 weakly excites
- L13:963 weakly excites
- L15:1541 weakly excites
- L21:2809 weakly suppresses
- L11:2638 weakly excites
- L14:607 weakly excites
- L22:1240 weakly excites
- L13:2668 weakly excites
- L18:1507 weakly excites
- L12:2024 weakly excites
- L14:1376 weakly suppresses
- L16:445 weakly excites
- L11:1848 weakly suppresses
- L12:594 weakly excites
- L16:1917 weakly suppresses
- L15:434 weakly excites
- L24:1896 weakly excites
- L17:2536 weakly excites
- L11:2849 weakly suppresses
- L21:2934 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
- whel
- ieg
- coaster
- โฮ
- 之余
- ören
- 衷
- OTS
- 抗衡
- .sections
Suppresses
- mong
- bson
- erra
- ubar
- lump
- .sav
- unic
- vascular
- 氓
- onView