space token right after a period
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. 0 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly 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 fires on the blank space that appears right after a full stop, mostly in citations, reference lists and cut-up text. It is a weak and fairly unclear signal. When it was turned up, the replies did not show a clear shared theme.
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.
-
found to be “obstructive.”<|endoftext|>Rural earthen roads impact assessment in Phewa watershed, Western region, Nepal⏎© The Author(s). 2016⏎Received: 23 November 2015
-
. 4- The subject becomes object and use (by) before it.⏎SubjectObject
-
. 2005. Edge Effects in Line Intersect Sampling With. Journal Agricultural
-
. 29:1–6)||Rosh Hashanah (head<|endoftext|>T
-
. 25:24–25.)⏎The law leads us back home;
-
. ¼ mile away). Touch for a list and map of all markers in Georgetown.⏎
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: to get the to get the to get the to get the to get the to get the to get the to get the to get the I have been thinking about lately: to get the to
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
.…It's less powerful than even a small…watch battery.…For processing and mixing
-
and his followers, social capital is a key component to building and maintaining democracy.⏎Communitarians seek to bolster social capital and the institutions of civil society. . . .⏎Central to many communitarians' philosophy is the concept of positive rights;
-
. Armed with the knowledge of where<|endoftext|>Edited by William Byron Forbush This is a book that will never die — one of the great English classics. . .
-
next-generation system was twice as much as it originally calculated, so it was forced to shut down the Arizona pilot and lay off nearly half of its staff.⏎. . .⏎So far so good. Those who think the algae are done are in for
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
None: this is the first layer.
Feeds later layers
- L1:130 excites
- L1:441 weakly excites
- L4:295 weakly excites
- L16:432 weakly excites period after month or initial
- L13:148 weakly excites
- L1:1071 weakly excites period inside abbreviations like i.e. and e.g
- L1:97 weakly excites full stop ending a heading or label
- L2:156 weakly excites
- L2:1243 weakly suppresses
- L1:1054 weakly excites "especially" after a comma
- L18:713 weakly excites
- L13:147 weakly excites
- L16:1802 weakly suppresses
- L15:313 weakly excites decimal point inside numbers
- L7:1760 weakly suppresses
- L1:448 weakly suppresses
- L3:595 weakly excites
- L1:1211 weakly suppresses
- L7:1219 weakly suppresses
- L3:754 weakly suppresses
- L17:2547 weakly suppresses
- L1:443 weakly suppresses
- L5:1329 weakly excites
- L2:1205 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
- œur
- ctest
- ives
- пуска
- come
- levant
- 不同
- ugins
- ʎ
- 汐
Suppresses
- sov
- Kont
- ItemType
- 穑
- enschaft
- .Annotations
- ronics
- urg
- `%
- `_