word-start 'con' fragment
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. 26 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 reacts to the 'con' piece at the start of words like 'confluence', 'connotations' or 'conveys'. Its negative side is unclear, and turning it up or down gave loops that do not clearly relate to this pattern.
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.
-
an appointment online at www.MyFortMyersDentist.com.<|endoftext|>In a world where 1 in 7 people are displaced, your kindness is just condescension.⏎The pirogues are long and narrow, precision-chise
-
and lose meaning or beauty.⏎If you love the poem a whole lot, spending more time and attention on it through memorization just creates additional lovable meanings, connotations, and associations. The poem becomes deeper and richer instead of falling apart.⏎
-
ative” (signed “story”) is sometimes preferred over “oral” because of an enduring history of “oral” education that has powerful, often painful, connotations for many Deaf adults. Historically “oral” education meant learning without
-
ies, and were soundly defeated in a battle in the western delta.⏎After two years of peace, another, more dangerous coalition, the Sea People, a conglomeration of migrating peoples from Asia Minor and the Mediterranean islands who had previously
-
field, was the birthplace of modern Manchester.⏎It was established by 78 AD at a site overlooking the confluence of the rivers Irwell and Medlock.⏎This was a timber fort with
-
centered and bathed in sunlight, has become covered in flour while he hurriedly scoops up every last grain from a looted depot. The photograph powerfully conveys the animalistic condition of the city's population in the aftermath of the conflict
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: willververververververververververververververververververververververververververververververververververververververververververververververver
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
40,000 British, Dutch and German troops 250 miles (400km) in five weeks. This extraordinary feat of strategy and logistics left Marlborough's opponents in disarray. Marlborough's tactics included building
-
likes” and Artificial Intelligence (AI) algorithms, machines can accurately predict personality traits such as openness, agreeableness, extraversion, conscientiousness, and neuroticism. He further states that AI algorithms, combined with a small number of
-
the journey with Piet, past soldiers and enemies, is fraught with danger. Along the treacherous path to Belgium the three children skate using every bit of speed, courage, and strength they can muster. All the time they
-
similar to the ruby-throats but are slightly larger and hardier and are differentiated primarily by their range. The broad-tailed<|endoftext|>A woman of extra-ordinary beauty, born in the Old Bailey. She⏎was twelve years a courtes
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
- L17:525 suppresses
- L25:1807 weakly suppresses
- L18:80 weakly excites
- L15:2250 weakly excites
- L6:362 weakly excites
- L2:2797 weakly suppresses
- L24:1553 weakly excites
- L2:1808 weakly excites
- L5:78 weakly suppresses
- L0:1575 weakly suppresses
- L16:1890 weakly suppresses
- L19:1385 weakly excites
- L19:850 weakly suppresses
- L7:43 weakly suppresses
- L21:1886 weakly suppresses
- L17:219 weakly excites
- L23:2881 weakly excites
- L10:3036 weakly excites
- L22:1273 weakly excites
- L14:2928 weakly excites
- L6:1414 weakly excites
- L4:2290 weakly suppresses
- L16:84 weakly excites
- L3:969 weakly excites
Feeds later layers
- L27:1015 weakly suppresses
- L27:1768 weakly excites
- L27:2750 weakly suppresses
- L27:2125 weakly excites
- L27:2562 weakly suppresses
- L27:1631 weakly suppresses
- L27:1539 weakly suppresses
- L27:2081 weakly suppresses
- L27:2874 weakly suppresses
- L27:1280 weakly suppresses
- L27:2615 weakly excites
- L27:2975 weakly suppresses
- L27:2571 weakly excites
- L27:674 weakly suppresses
- L27:544 weakly suppresses
- L27:2303 weakly suppresses
- L27:1471 weakly suppresses
- L27:2581 weakly suppresses
- L27:1749 weakly excites hyphen joining word parts
- L27:1834 weakly suppresses
- L27:2629 weakly excites
- L27:986 weakly suppresses
- L27:2469 weakly suppresses
- L27:2595 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
- bservable
- ично
- но
- igrations
- terior
- uppy
- pecia
- orda
- erno
- ounds
Suppresses
- yen
- ')</
- 一角
- ayar
- �
- YW
- -END
- сте
- ver
- .Pages