'in' after 'high' or 'dressed'
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. 8 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 "in" when it follows words like "high" or "dressed", as in "high in vitamin A" or "dressed in white". The same word "in" in other places, such as "floats freely in space", pushes the neuron the other way. Turning it up did not give a clear theme in the model's writing.
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.
-
act as a natural sleep aid.⏎Pumpkin is high in vitamin A and carotenoids – beta carotene, lutein,
-
Europe and the Middle East and Central Asia for longer than many people realize.⏎If you walked behind the owner of this skull when he was alive and he was dressed in a modern outfit, you would think that he was just another human being until you
-
as Lucia described:⏎When Our Lady disappeared in the immense distance of the sky, next to the sun we saw Saint Joseph holding the Child Jesus and Our Lady dressed in white with a blue mantle. Saint Joseph and the Child seemed to be blessing the
-
V.⏎This<|endoftext|>Warning: Use of undefined constant user_level - assumed 'user_level' (this will throw an Error in a future version of PHP) in /home/commons/public_html/wp-content/plugins/ultimate-google-analytics/ultimate_ga.php
-
of cancer, including lung, stomach, prostate, cervical and breast cancer.⏎2. Supports your Heart Health⏎Because these nuts are high in monounsaturated fats, they are able to reduce the bad cholesterol, and also
-
of linoleic acid and 0.432 g of Linolenic acid.⏎Nutritional facts Per 100 Grams⏎Helps in improving vision⏎Beta-carotene is very much important for improving vision. Many
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 not a place where I have been for a long time. I am going to be a place where I am for the long term. I am a place that is meant to be a position that is a certain way. The way
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
1.4 million light-years away from Earth, so far away that it floats freely in space, unperturbed by the Milky Way.⏎Unlike the other hobbit galaxies
-
Currents⏎Generation of Alternating Voltages and Currents⏎An alternating voltage may be generated⏎(1) by rotating a coil at constant angular velocity in a uniform magnetic field as shown in Fig. (a).⏎(2) by rotating
-
an athlete jumped 2 feet high and left the ground at an angle of 20 degrees with respect to the horizontal, how fast was the athlete going in the forward (positive horizontal) and upward (positive vertical) directions immediately after takeoff
-
tightly constructed plot.⏎Carol Milford Kennicott, a graduate of "sanctimonious" Blodgett College, with a year of additional study in a Chicago library school, works as a librarian in St. Paul
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
- L4:289 excites
- L5:490 excites
- L3:493 suppresses
- L7:2824 suppresses
- L4:1966 suppresses
- L2:2307 weakly excites
- L0:781 weakly suppresses
- L6:2937 weakly suppresses 'by' after a title or author credit
- L4:430 weakly excites
- L3:3065 weakly excites
- L3:1583 weakly excites
- L5:910 weakly suppresses
- L2:963 weakly excites
- L4:2389 weakly suppresses
- L1:1895 weakly excites
- L4:2642 weakly excites
- L2:2824 weakly excites
- L1:3004 weakly excites
- L3:1200 weakly suppresses
- L4:1226 weakly excites
- L4:1834 weakly excites unclear: time spans and 'once you've'
- L1:2621 weakly excites
- L3:1489 weakly suppresses
- L4:1639 weakly excites
Feeds later layers
- L16:1028 suppresses "in" after a superlative or ordinal
- L19:649 suppresses
- L21:1935 suppresses
- L16:1557 suppresses 'in' meaning 'in this case' or 'in between'
- L23:1366 suppresses
- L25:1865 suppresses 'in' in 'keep in mind'
- L19:1345 suppresses
- L17:165 suppresses
- L19:704 suppresses
- L17:2609 suppresses
- L23:967 suppresses
- L18:1588 suppresses
- L17:2126 suppresses
- L15:871 suppresses
- L17:706 suppresses
- L16:2892 suppresses
- L25:2957 suppresses 'in' starting a word fragment
- L23:2136 suppresses 'in' after a noun phrase
- L16:1807 suppresses 'in' after a motion verb
- L12:1066 excites
- L24:2155 suppresses
- L9:3031 suppresses
- L16:2480 suppresses
- L11:1187 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
- 自动生成
- рядом
- Ire
- addChild
- onView
- ро
- icularly
- requests
- requestData
- addChild
Suppresses
- context
- contexts
- areas
- jur
- case
- terms
- dont
- direction
- environments
- land