'con' starting a word piece
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. 27 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 syllable 'con' when it begins a word fragment, as in constrict, congregate, concomitant and condescended. Turning it down made the model output a stream of letter c's, and turning it up made it repeat words like 'finals' and 'whether'.
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.
-
brain’s reward center. It can also cause blood vessels to constrict and muscle relaxation. It may also impair memory and concentration. Nicotine can
-
a major, preventable cause of harm. To improve knowledge on anticoagulant safety, we plan to conduct bidirectional self-controlled case series studies, among concomitant users of warfarin and a skeletal muscle relaxant and, as
-
organic methods, with no chemical pesticides or fertilisers,” Crocco told IPS.⏎The crops planted in the school gardens, depending on the season, include lettuce, arugula, Swiss chard, spinach, bell peppers, tomatoes, squash,
-
every night, while others snore only occasionally.⏎What is sleep apnea?⏎Sleep apnea is similar to snoring, except that the airway becomes completely constricted. As a result, it becomes impossible to breathe. The typical pattern for
-
had a request to share more about that last one- so we’re meeting some snakes from around the world.⏎Today we’re meeting the black rat snake- a constrictor and helpful snake.⏎Black Rat Fun Facts:⏎- They are also known
-
not elective or historical. They reach right into those classrooms, affecting every single student. Students we worked with talked of being cat-called, harassed, objectified, condescended to and, most notably, excluded from the same spaces their male counterparts
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: Cc 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 c
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
known to result in appetite loss. Plus, the pain, fatigue and other<|endoftext|>This Super food is good source of fiber and has only 50 calories per one-cup serving⏎Pumpkin can keep you feeling full longer
-
) and various other nutrients.⏎Protein content: 59% of calories. One cup (226 g) of cottage cheese with <|endoftext|>National Mammal Week
-
on fewer calories. Why? Because pumpkin is a good source of fiber and has only 50 calories per one-cup serving of mashed pumpkin. Recent research is even finding that pumpkin can help
-
oids, B vitamins, and fiber. It is also low in calories.⏎Nutritional Value and Health Effects⏎A 100-gram (3.5 oz) serving of oca provides almost 40 milligrams of vitamin C
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
- L6:1963 weakly suppresses
- L3:576 weakly suppresses
- L4:2 weakly suppresses
- L26:3004 weakly suppresses
- L24:105 weakly suppresses
- L14:1627 weakly suppresses
- L7:102 weakly suppresses
- L3:59 weakly excites
- L21:237 weakly suppresses
- L0:1057 weakly excites
- L26:1564 weakly excites
- L23:206 weakly suppresses
- L18:1083 weakly suppresses
- L22:2365 weakly suppresses
- L26:581 weakly suppresses
- L15:2447 weakly suppresses
- L2:761 weakly excites
- L1:23 weakly suppresses
- L1:2677 weakly suppresses
- L14:2441 weakly suppresses
- L0:2701 weakly suppresses
- L21:236 weakly excites
- L16:1507 weakly excites
- L23:1685 weakly excites
Feeds later layers
None: this is the last layer.
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
- 姊妹
- bih
- eses
- @stop
- 因为它
- crest
- _EDEFAULT
- Footer
- prises
- ⎦
Suppresses
- c
- c
- C
- C
- _c
- .c
- *c
- ;c
- cac
- →c