angle and right-angle 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. 20 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 lights up on the word 'at' or 'a' right before an angle is named, like 'at right angles' or 'at a slight angle'. When it was turned up, the model kept bringing up angles, even on unrelated topics like the weather.
Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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.
-
defined in terms of its⏎(i) Isometric System :⏎Three axes of crystal are⏎of equal length and lie at right angle to each⏎other, e.g. Halite mineral.⏎(ii)
-
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 and if
-
ai tree needs watering.⏎Take your bonsai tree outdoors and set it on a flat surface. Fill a watering can with water and tip the spout downward at a slight angle to water the container slowly and
-
Vancouver.<|endoftext|>1. The problem statement, all variables and given/known data If 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
-
fridge.⏎To see this for yourself, stick two fridge magnets together, back to back. They will slide against each other in one direction. If you slide them at right angles to that, they will alternately attract and repel one another.
-
Dulles International Airport and flying a Boeing 727 due east at 450 mph. There is a 65 mph wind with the bearing 60 degrees.⏎a. Determine the compass heading that Reem should follow.⏎
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 current system of pressure management (P2P) for the last two years have been so far a lot of negative pressure and a lot of lack of work, and the latest new P2P is the 'last year's P2
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
(theta) Immediately after take off : vx = 2cos25 = 1.8126 vy = 2sin25 = 0.845 right before landing: vx = 2cos25 =
-
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
-
pet owners will agree wholeheartedly that 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
-
9.6 km/h)|⏎|Range:||3,500 nautical miles (6,482 km) at 12 knots (22.2
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
- L9:2706 excites
- L11:2222 excites
- L10:1234 suppresses
- L9:2533 excites
- L15:2072 excites
- L18:2426 suppresses
- L10:825 suppresses 'B' in vitamin B names
- L7:1328 suppresses
- L17:2847 suppresses
- L18:569 excites
- L13:2876 excites
- L9:1602 suppresses
- L3:366 excites
- L8:3047 weakly excites
- L17:1573 weakly excites
- L3:2081 weakly suppresses
- L10:2577 weakly suppresses
- L7:2066 weakly suppresses unclear: technical word fragments, mostly engineering
- L18:938 weakly suppresses
- L6:1818 weakly suppresses
- L6:1969 weakly excites
- L8:2820 weakly excites
- L2:1334 weakly suppresses
- L9:1769 weakly excites
Feeds later layers
- L23:2260 strongly excites
- L22:1574 strongly excites
- L21:2787 excites
- L27:2041 excites
- L23:2686 excites
- L24:2883 excites
- L25:3034 excites
- L21:2220 excites
- L22:2564 excites perpendicular and geometry terms
- L22:2165 weakly suppresses
- L21:1203 weakly suppresses
- L23:2142 weakly excites
- L21:2768 weakly excites
- L21:1566 weakly suppresses
- L23:1474 weakly suppresses
- L23:1126 weakly excites
- L25:1119 weakly excites
- L21:1283 weakly suppresses
- L21:1897 weakly excites
- L22:1432 weakly excites
- L23:669 weakly excites
- L24:2954 weakly suppresses
- L23:1894 weakly excites
- L25:1116 weakly excites
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
- angle
- 角度
- 的角度
- angles
- Angle
- Angle
- _angle
- góc
- angle
- -angle
Suppresses
- 墟
- press
- .sel
- 公报
- ppelin
- .press
- نسخ
- elijk
- 昼
- 已久