comparison targets like 'than Earth'
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. 14 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 the thing being compared to, such as 'Earth' in 'larger than Earth' or 'much like Earth'. Turning it up gave text that did not obviously relate to comparisons.
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.
-
with a venting parent. As a result, they might have learned to suppress their anger for fear of retaliation or made a personal vow never to be like their parent.⏎The venter style is at the opposite end of the anger management continuum.
-
exoplanets — could also be rocky and have a composition much like Earth’s. Some might even harbor life. Over 99% of those alien
-
coil that induces them to combine into helium atoms. The process also releases neutrons.⏎The Fusion Engine provides energy in two ways. Like in a fission reactor, the energy of the scattered neutrons gives off heat that ultimately drives a turbine.
-
, built by Airbus, of methane and carbon dioxide in the atmosphere of K2-18b, an exoplanet 8.6 times bigger than Earth, shows there is still a lot to uncover in the search for habitable environments
-
determined autocrat but not an antisemite. Portugal’s very small and then only recently reestablished Jewish community possessed the same rights as the country’s other citizens. It retained
-
to discover planets ranging from huge gas giants, some of them orbiting close to their stars, to “super Earths,” Earth-like planets that are larger than Earth but still might contain life.<|endoftext|>New guidelines have been launched for educators, using
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: something that has to do with something that has to do with something else... something... something... something... what... what... what... what... what... what... what... what... what... what... what... what... what... what
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
striking; certain passages of the hadeeth could easily be mistaken for discussions about contemporary environmental issues.⏎The Prophet sall Allaahu`alayhi wa sallam's environmental philosophy is first of all holistic: it assumes a fundamental link and inter
-
GNU package:octave⏎released on 31 May 2012⏎|License||Verified by||Verified on||Notes|⏎|GPLv2||Kelly Hopkins||6 December 2011|⏎|GPLv3
-
ious balance between man and nature. From all accounts of his life and deeds, we read that the Prophet sall Allaahu`alayhi wa sallam had a profound respect for fauna and flora, as well as an almost visceral connection
-
to teach you to observe in the land that you are crossing the Jordan to possess, so that you, your children and their children after them may fear the Lord your God as long as you live by keeping all his decrees and commands that I
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
- L6:982 excites
- L3:577 excites
- L10:2125 suppresses
- L2:291 weakly suppresses
- L12:2772 weakly suppresses comma after an introductory phrase
- L2:937 weakly suppresses
- L7:2418 weakly suppresses
- L11:708 weakly suppresses
- L12:625 weakly excites
- L8:1259 weakly suppresses
- L11:525 weakly suppresses parentheses and code fragments
- L8:971 weakly suppresses
- L3:2884 weakly suppresses
- L12:1021 weakly suppresses
- L11:2214 weakly excites
- L11:2568 weakly excites
- L12:1040 weakly excites
- L4:2499 weakly suppresses
- L8:868 weakly excites
- L13:528 weakly excites
- L9:1945 weakly suppresses
- L5:2896 weakly suppresses
- L8:316 weakly excites
- L1:157 weakly suppresses
Feeds later layers
- L17:2737 weakly excites
- L22:702 weakly excites
- L18:1822 weakly excites
- L19:2955 weakly excites
- L15:452 weakly excites
- L24:439 weakly excites
- L21:2726 weakly excites
- L17:1942 weakly suppresses
- L24:532 weakly excites
- L18:2303 weakly excites
- L15:401 weakly excites
- L21:1041 weakly suppresses
- L15:1470 weakly excites
- L18:2441 weakly suppresses
- L21:2046 weakly excites
- L20:246 weakly excites
- L25:1172 weakly excites
- L18:2631 weakly excites
- L23:1334 weakly excites
- L16:789 weakly suppresses
- L21:1405 weakly excites
- L18:2117 weakly excites
- L15:2956 weakly excites
- L18:2459 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
- explicit
- 隅
- formance
- 舞台上
- 奔跑
- (Key
- 适时
- 早点
- Am
- ListAdapter
Suppresses
- grown
- 系
- essel
- 故
- itus
- inent
- "-//
- i
- 洗净
- uelle