period in a middle initial
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. 7 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 full stop after a person's middle initial, like the dot in 'William R. Pogue'. The activations are fairly weak, and turning it up gave text that does not obviously relate to names.
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.
-
table, ' get its DataSet. Dim thisDataSet As DataSet = CType(DataGrid1.DataSource, DataSet) ' Get the DataTableCollection. Dim tablesCollection As DataTableCollection = thisDataSet.Tables ' Get a specific table by name. Dim table As DataTable
-
The Mercury 7 Astronauts⏎William R. Pogue orbited the Earth for 84 days during the final manned flight
-
Dakota Game, Fish and Parks Department Authors are James R. Johnson and Gary E. Larson⏎$19.95 and Free Shipping
-
The President of the United States has seven roles in his job, from being the Commander-In-Chief to the Guardian of the Economy.⏎Brief biography of Gerald R. Ford, the 38th president of the United States of America.⏎Timeline
-
While taking some time in the middle to form the Industrial Workers of the World. Debs died in 1926.⏎5: Sources " Eugene V. Debs | Biography
-
to Thomas Aquinas, to Mortimer J. Adler. Now Adler instructs the area within the "uncommon good judgment"
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: the "s" and "r" in the game of "Sudoku". So, I'm going to take the lead with a new game. I have a new puzzle board, and I want to play. Now, I'm going
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Mar. 1, 2007).⏎In 2006, the Department of Homeland Security (DHS)’s Customs and Border Protection apprehended 101,952 unaccompanied children (ali
-
can now be a threat to world peace. Finally, the U.S. government decision to adopt unilateral measures, bypassing the United Nations and the International Court of Justice, has jeopardized the slow, tentative process of creating a
-
promoting the rule of law, ensuring access to justice, and strengthening institutions. It is the intersection of SDG 11 and SDG 16 that holds importance, and hope, for vulnerable populations in fragile communities around the world
-
waters. Last year both governments signed a revised Great Lakes Water Quality Agreement and once again asked the IJC to report regularly on the health of the
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
- L6:16 weakly excites
- L6:13 weakly excites
- L4:1308 weakly excites
- L3:2492 weakly suppresses
- L6:2834 weakly excites
- L2:923 weakly excites second half of 'side by side' and 'face-to-face'
- L3:1561 weakly excites
- L6:1970 weakly suppresses
- L1:1687 weakly excites
- L6:8 weakly suppresses
- L2:1490 weakly excites
- L5:1194 weakly suppresses
- L6:1520 weakly suppresses
- L3:1431 weakly excites
- L3:512 weakly suppresses
- L1:1441 weakly suppresses
- L3:2732 weakly excites
- L6:12 weakly excites
- L1:1343 weakly suppresses
- L4:99 weakly excites
- L5:742 weakly excites 'this' after a preposition
- L0:1808 weakly excites
- L4:21 weakly excites
- L2:2710 weakly excites
Feeds later layers
- L8:446 excites
- L10:1241 excites
- L12:2009 excites
- L20:900 excites end-of-text right after cut-off text
- L8:224 suppresses
- L11:1181 excites
- L9:853 weakly suppresses
- L8:121 weakly excites
- L8:2187 weakly suppresses
- L12:2131 weakly suppresses
- L19:619 weakly excites
- L12:246 weakly excites
- L18:151 weakly excites
- L18:195 weakly excites
- L8:1917 weakly excites
- L8:1998 weakly excites
- L16:2339 weakly suppresses
- L19:2146 weakly suppresses
- L16:2724 weakly excites
- L21:1337 weakly suppresses
- L8:530 weakly suppresses
- L13:450 weakly excites
- L14:1847 weakly excites
- L18:2909 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
- 因为你
- ue
- eval
- lest
- uen
- kj
- acionales
- 竞
- ::$
- ecz
Suppresses
- -dot
- 딩
- arding
- par
- átis
- 墉
- ERGY
- ptime
- romium
- .Has