"to" in "click here to"
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. 3 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 word "to" in the phrase "Click here to...", the kind of link text found on web pages. Turning it up made the writing wander into unrelated topics, so the steering doesn't show this pattern clearly.
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.
-
from a traffic light post in the streets of the capital.⏎The contents of this article is licensed from www.wikipedia.org under the GNU Free Documentation License. Click here to see the transparent copy and copyright details<|endoftext|>What is Celiac / Coeliac Disease
-
Kings, Revolutionaries, and Statesmen who fought to free this land; it is an attempt to kindle the spirit of patriotism in the youth !⏎Click here to view Gallery⏎Let their
-
processor.⏎What is the prerequisite to learn PHP?⏎- If you want to use PHP to create web sites and want to write code for client-side (Click here to know more about Client-side language), then you need to know HTML before learning PHP
-
child's development and, on the 'Words for Life' website, there are some useful tips about how you can help your child when doing this. Click here to read these. The website also provides tips about sharing stories with your child which can
-
izing your cable entertainment bill over protecting and investing in your family?⏎Smart Policies are as low as $30 a month, No Medical Exam Required⏎Click Here to Get Smart on Protecting Your Family and Loves Ones, No Matter What Happens
-
learn to care for their child's needs.⏎The Osteogenesis Imperfecta Society can be an important resource for parents of children with OI.⏎Click here to view the⏎Online Resources of Orthopaedics<|endoftext|>
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: I am going to be in the process of the number, and I am going to be in the process of the number. That is the first, and the second, and that is the second. This is a very important number to me
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
99 [ flag... ] file... -lm [ library... ] #include <math.h> double ldexp(double x, int exp);⏎float ldexpf(float x, int exp);⏎long double ldexpl(long double x, int exp
-
[Filters->Renders->Clouds->Solid Noise] and give it these settings:⏎X Size: 6.0⏎Y Size: 4.0⏎Randomize the seed or find one you like.⏎If you want an ocean
-
blood cells⏎Plasmolysis of human red blood cells would occur if the cell were in .. solution A) isotonic B) hypertonic C) hypotonic D) none of these I think it's B. I'm not too
-
produce neonatal encephalopathy in⏎boys. The encephalopathy is characterized by seizures, cortical myoclonus,⏎hypotonia and central apnea. Rett syndromes in male neonates may
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
- L0:1224 suppresses
- L0:178 suppresses
- L0:1301 suppresses
- L0:2064 weakly suppresses
- L2:126 weakly excites
- L2:49 weakly suppresses
- L1:1114 weakly excites
- L0:518 weakly suppresses
- L0:1061 weakly excites
- L0:2586 weakly excites
- L2:571 weakly suppresses
- L2:156 weakly suppresses
- L1:1622 weakly suppresses
- L0:1639 weakly suppresses
- L0:1278 weakly suppresses
- L1:2547 weakly suppresses
- L2:75 weakly suppresses
- L1:1464 weakly excites
- L1:2946 weakly suppresses
- L1:2014 weakly excites
- L2:2535 weakly suppresses
- L0:557 weakly excites
- L1:2058 weakly excites
- L1:2801 weakly excites
Feeds later layers
- L4:1831 suppresses
- L4:385 suppresses multiplication and scaling in math
- L4:1 weakly suppresses
- L5:2079 weakly suppresses
- L12:2685 weakly excites
- L12:2180 weakly excites
- L14:1559 weakly suppresses
- L6:1351 weakly suppresses
- L22:2279 weakly suppresses
- L11:1752 weakly excites
- L5:1559 weakly suppresses
- L5:211 weakly suppresses
- L14:1547 weakly excites
- L17:2465 weakly excites
- L22:465 weakly excites
- L15:749 weakly excites
- L12:895 weakly suppresses
- L7:2675 weakly excites
- L4:1696 weakly suppresses
- L12:714 weakly suppresses
- L21:2872 weakly suppresses
- L15:726 weakly suppresses
- L18:1200 weakly excites
- L5:1081 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
- ethe
- ez
- -sh
- /themes
- hend
- quare
- 直辖
- ControlEvents
- ihar
- ele
Suppresses
- .assertAlmostEqual
- diffs
- inaire
- isNaN
- 능
- illo
- diff
- →diff
- rance
- auprès