'to' and 'of' inside phrases
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 fairly consistent: several of them 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 responds to small connecting words like 'to' and 'of' that sit in the middle of a phrase, often after a verb or noun. It goes quiet, or slightly negative, on fragments of longer words such as the end of 'alleys'. Turning it up or down did not change the writing in a way that matches this pattern.
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.
-
,<|endoftext|>This material must not be used for commercial purposes, or in any hospital or medical facility. Failure to comply may result in legal action.⏎WHAT YOU NEED TO KNOW:⏎Diphtheria is a disease caused by a bacterial infection. The
-
technique in the opening scene, when Enfield relates his story of witnessing Hyde trample a little girl—a night when the streets were so empty that he began “to long for the sight of a policeman.”⏎This notion of the city as a fearful
-
HAVE DIFFERENT REACTION RATES BECAUSE OF DIFFERENT STRUCTURE OF THE COMPOUNDS. DIFFERENT REACTION WILL YIELDS TO A CONCLUSION OF WHAT IS/ARE THE FACTOR/S OF ELECTROPH
-
two component molecules, and will⏎reassociate upon slow cooling under appropriate salt conditions. So,⏎the issue is not decomposition so much as disassociation upon exposure⏎to too high a temperature; and the reason for the disassociation is that⏎the
-
genes has led to the identification of highly conserved consensus sequences⏎at the 5' and 3' ends of essentially all mRNA introns.(Figure⏎of Consensus Splice Sites) A majority of these searches are based on algorithms that
-
⏎the answer may not have known in fact if someone else had been naturalized.⏎An individual may have said yes because he felt it was the right thing⏎to say or he intended to begin the process.⏎-
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: la question de liserie, le liserie, le liserie, la liserie, la liserie, la liserie, la liserie, la liserie, la liserie, la
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
promoters, Ace Borger and J.T. Peyton, were engaged. The land was then surveyed, subdivided into blocks, lots, streets, avenues, and alleys. The Stinnett Town site Company dedicated the streets, avenues and alle
-
Rights Division⏎A Guide To Federal Religious Land Use Protections⏎The Religious Land Use and Institutionalized Persons Act (RLUIPA)⏎protects religious institutions from und
-
of errors in the street grid, labeling Locust Street as “Lock Street” and Grand Avenue as “Main Street.” Nor does he depict the sixteen-foot-wide alleys that are clearly platted on the Sanborn maps, although in several cases
-
linear, curving hills and depressions, which are most distinct in the right of the image. Many narrow, linear grooves run in various directions across this undulating terrain. There are some small, less than 1 kilometer (0
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
- L1:341 suppresses
- L3:7 suppresses
- L2:21 excites
- L1:43 weakly excites
- L1:571 weakly suppresses
- L16:253 weakly suppresses
- L1:150 weakly excites
- L2:17 weakly excites
- L17:385 weakly excites
- L17:852 weakly excites
- L16:890 weakly suppresses
- L2:407 weakly excites
- L1:862 weakly excites
- L17:2014 weakly excites
- L22:2455 weakly excites
- L23:2691 weakly excites
- L3:504 weakly excites
- L23:312 weakly suppresses
- L26:2676 weakly suppresses
- L1:232 weakly excites
- L10:301 weakly excites
- L17:707 weakly suppresses
- L19:1020 weakly suppresses
- L3:722 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
- 菟
- 恚
- well
- ięć
- perfection
- _cookies
- TERS
- tep
- numbering
- 完美
Suppresses
- ingly
- 主人
- <>⏎
- /task
- /query
- quina
- /event
- halb
- ös
- ogenous