perpendicular and geometry terms
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. 22 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 the word 'perpendicular' and to nearby geometry vocabulary like lines, angles and slopes. When it was turned down, the model's answers drifted toward maths-style wording, though the links are loose.
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.
-
circle touches the larger circle and. . . .⏎X is a moveable point on the hypotenuse, and P and Q are the feet⏎of the perpendiculars from X to the sides of a right angled⏎triangle. What position of
-
's square.⏎Find the area of the shaded region created by the two overlapping⏎triangles in terms of a and b?⏎Position the lines so that they are perpendicular to each other.⏎What can you say about the equations of perpendicular lines?⏎AB
-
are parallel, perpendicular or neither?⏎Parallel lines are the lines that never intersect and perpendicular lines are the lines that intersect at 90 degrees.⏎Answer: Lines with the same slope are parallel
-
slopes are m = 2 and m = –1/2⏎If the slope of one line is the negative reciprocal of the second line, then lines are perpendicular.⏎Hence, both the lines are perpendicular.
-
two perpendicular components namely, IRMS cos φ along VRMS and IRMS sin φ perpendicular
-
relieve the symptoms of the early response, but they have little effect on late stage.⏎The non-bronchodilating or anti-inflammatory drugs such as cromolyn (Intal) and cortisone do not relieve symptoms once they occur
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 sidec c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
their pets are loving and unselfish! Share this finding with your dog-loving friends!<|endoftext|>How to Graph a Cotangent Function⏎The parent graphs of tangent and cotangent are comparable because they both have asymptotes and x-intercepts. The
-
parent graph of the cotangent function f(x) = cot x,⏎by using the same techniques you use to find the tangent parent graph
-
, or isosceles.⏎a. A triangle with no congruent sides is called __?
-
the expressions of V(t), I(t), the trigonometric formula sin(X)sin(Y)=1/2(cos(X-Y)-cos(X+Y)), and the fact that (Vmax
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
- L20:1515 excites
- L6:946 suppresses
- L20:2626 excites angle and right-angle phrases
- L10:1571 suppresses
- L8:2674 weakly excites
- L21:2877 weakly suppresses
- L9:2117 weakly suppresses
- L17:2890 weakly excites
- L18:408 weakly excites
- L5:2452 weakly suppresses
- L6:779 weakly suppresses
- L3:187 weakly excites
- L11:1909 weakly excites
- L14:2952 weakly excites
- L20:1438 weakly suppresses
- L21:2408 weakly suppresses
- L9:1776 weakly suppresses
- L6:1814 weakly suppresses
- L4:2016 weakly excites
- L4:660 weakly excites
- L11:2695 weakly excites energy and resource nouns
- L17:488 weakly suppresses
- L4:540 weakly suppresses
- L10:2411 weakly excites
Feeds later layers
- L26:1055 suppresses
- L25:1466 suppresses
- L23:333 weakly suppresses
- L23:2554 weakly excites
- L27:495 weakly suppresses
- L24:676 weakly suppresses
- L24:502 weakly suppresses
- L23:2459 weakly suppresses
- L27:1833 weakly suppresses
- L26:1146 weakly suppresses
- L26:1427 weakly suppresses
- L23:801 weakly suppresses
- L23:450 weakly suppresses unclear: raccoon name fragments
- L24:1626 weakly suppresses
- L24:2393 weakly suppresses
- L24:714 weakly suppresses
- L24:1869 weakly suppresses
- L25:302 weakly suppresses
- L24:2622 weakly excites
- L23:941 weakly suppresses
- L23:741 weakly suppresses
- L25:3011 weakly suppresses
- L25:1154 weakly suppresses
- L27:917 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
- mdi
- .MESSAGE
- ylene
- rale
- ories
- 瓦
- 凰
- /mark
- ranking
- 皱纹
Suppresses
- cott
- c
- side
- sides
- )c
- _side
- сон
- ไซ
- side
- SIDE