'followed' and 'searched' after a subject
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. 18 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 verbs like 'followed', 'following', 'searched' and 'dreaming' that come right after a subject, as in 'scientists had searched'. Its strongest negative cases are truncation markers like '[...]' or '…' that cut a sentence short. Turning it up made the model drift into abstract, vague statements.
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.
-
. Only 183 seconds long, the simple melody and poetry captured the wounded hopefulness of its moment - and transcended its time to inspire generations that followed.⏎Charles J. Shields traces the song's origins - from the fire-bomb
-
Do not look directly at oncoming headlights, rather look toward the right side of the road and use the white fog line as a guide. When headlights from vehicles following you reflect in your rearview mirror, use the “day-night” feature on
-
ela, a professor at the University of California, Santa Barbara and the former spokesperson of the CMS experiment at the Large Hadron Collider.⏎For decades, scientists had searched for the elusive Higgs boson: the holy grail of modern physics and
-
was begun on a holiday at Braemar in Scotland at the prompting of Fanny’s young son. Kidnapped and The Master of Ballantrae followed on as his international fame as<|endoftext|>“Paradise Found and Lost” from
-
.⏎You may also like..⏎- Education and Training⏎- Special Populations⏎- Youth in Transition<|endoftext|>White Turkey – White Turkey⏎No, you’re not dreaming – White turkey are a real thing! White turkeys are bred from stock that
-
politics. The Ohio-born Hayes followed a fairly typical trajectory to the White House, checking off many of the most common
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: how to do a small ...??????????????????????????????????????????
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
aphylaxis-like symptoms previously should always have injectable epinephrine (an Epi-Pen) nearby.⏎This answer should not be considered medical advice...This answer should not be considered medical advice and should not take the place of a
-
rather than cell references) from the ANOVA output as well as the Excel T... function. (Do not use the output from the ANOVA Template.xlsx file
-
ayanova, the Murray B. Sachs Professor in the Department of Biomedical Engineering at The Johns Hopkins University Schools of Engineering and Medicine and the project’s principal investigator… Continue reading.⏎Natalia Trayanova, a professor
-
bites are not initially painful and can go unnoticed for hours or days. This allows bed bugs to withdraw human blood for up to 10 minutes with each feeding...One bed bug will usually take more than one bite. Once a bed bug inserts
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
- L15:1978 strongly suppresses
- L3:990 excites
- L16:64 excites
- L17:360 suppresses
- L16:60 suppresses
- L0:2889 suppresses
- L16:572 excites
- L16:493 weakly suppresses
- L16:155 weakly excites
- L16:110 weakly suppresses
- L0:2471 weakly suppresses
- L0:1964 weakly excites enzyme and protein biochemistry terms
- L16:215 weakly suppresses
- L10:273 weakly suppresses
- L6:944 weakly suppresses
- L16:83 weakly suppresses
- L17:765 weakly suppresses
- L16:162 weakly excites
- L0:2872 weakly suppresses
- L2:2028 weakly excites
- L16:75 weakly suppresses
- L16:260 weakly excites
- L1:24 weakly excites
- L2:2029 weakly suppresses
Feeds later layers
- L23:641 suppresses
- L20:653 excites the 9 in years like 19xx
- L27:325 suppresses
- L19:1004 excites
- L19:902 excites
- L26:362 weakly excites
- L20:2507 weakly excites
- L26:626 weakly suppresses
- L21:334 weakly excites
- L26:889 weakly excites
- L27:2834 weakly excites
- L22:384 weakly excites
- L26:791 weakly excites
- L26:2878 weakly suppresses
- L23:3011 weakly excites
- L19:572 weakly excites
- L20:1228 weakly excites
- L19:557 weakly excites
- L20:2475 weakly excites
- L27:2912 weakly excites
- L22:1084 weakly excites
- L23:270 weakly suppresses
- L26:2249 weakly excites
- L20:2782 weakly excites
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
- pu
- glitter
- bie
- قو
- grinder
- 赎
- Toe
- ilig
- atism
- lass
Suppresses
- 続きを読む
- edException
- ynchron
- clamation
- ../
- 征
- artial
- enko
- 心意
- ple