word after "due to" or "with"
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. 18 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 small word right after phrases like "Due to", "With" or "despite", such as "the" or "its". It seems to track the start of a noun phrase after a cause or manner phrase. Turning it up gave text that did not clearly reflect this.
Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens an email to the maker with the neuron already filled in).
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.
-
ages⏎Everyone wants to live in a peaceful community free from conflict. Is it conceivable? Yes, if everyone chooses to get along with one another.⏎Due to the fact that the older and younger generations were not raised in the same manner, there
-
download page where you can immediately download the file and save it to your computer. You will also receive an email containing a link to the resource.⏎* Due to the digital nature of this product. Hot Chocolate Education does NOT offer refunds. Once the
-
small as asteroids go. Because of that (and its current distance of 450 million kilometers) it's a bit faint,
-
outcomes can be translated into a more frequent occurrence in my classroom. With all the conversations we have been having in PLC and ILT about wanting that real world
-
ners, the European Coal and Steel Community, the European Atomic Energy Community, and the European Economic Community, were all established by the end of the decade. With their emphasis on economic integration, European states found themselves increasingly reliant on the US-led NATO
-
peia constellation. It was discovered by Charles Messier in 1774. M52 can be seen from Earth with binoculars.⏎Due to interstellar absorption of light, the distance to M 52 is uncertain,
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 ability to a) a) a) a) a) a) a) a) a) a) a) a) a) a) a) a) a) a) a) a) a) a) a
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
that these two are witnesses against Darnay. Nonetheless, Lucie's face radiates a compassion that awes the spectators.⏎The compassion in Lucie's face indicates that
-
Butler Plantation in MacIntosh County, Georgia, in 1865. Her parents were devout Episcopalians, and also instilled in their children a love of learning. Anna became a teacher, and eventually the only African American to
-
transform into anger or guilt. Experts suggest deep breathing and connecting with your body to release these negative emotions.⏎What Is Compassion?⏎Compassion is the feeling or showing of sympathy and concern for others, as well as the consequent
-
the information gathered by the instruments you designed. This information is pivotal in gaining a clearer picture of how the aircraft is functioning.⏎This is a job that requires an acute attention to detail. Test flights are
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
- L17:2108 excites
- L4:309 excites
- L17:1757 suppresses
- L3:1218 weakly suppresses
- L10:1855 weakly suppresses
- L17:2685 weakly suppresses
- L6:2718 weakly suppresses
- L6:2708 weakly suppresses
- L15:835 weakly excites
- L15:1703 weakly excites
- L9:2478 weakly excites
- L16:2963 weakly suppresses
- L6:2293 weakly suppresses 'in' or 'the' in 'up in the air'
- L15:666 weakly excites
- L16:573 weakly suppresses
- L15:2729 weakly excites
- L10:1839 weakly excites
- L11:2005 weakly excites
- L1:1251 weakly suppresses
- L10:777 weakly suppresses
- L4:882 weakly suppresses 'with' after provide or familiarize
- L9:1993 weakly excites
- L14:1530 weakly excites
- L16:1737 weakly excites
Feeds later layers
- L21:440 suppresses
- L21:1079 suppresses
- L23:661 suppresses
- L19:1030 excites
- L22:749 suppresses
- L25:1133 suppresses
- L20:803 weakly suppresses
- L19:2406 weakly suppresses
- L20:566 weakly excites
- L22:809 weakly suppresses
- L19:2453 weakly excites
- L22:2137 weakly excites
- L19:2397 weakly excites
- L26:1527 weakly suppresses
- L19:2396 weakly suppresses
- L20:1596 weakly suppresses
- L20:341 weakly suppresses
- L21:1987 weakly suppresses
- L26:456 weakly excites
- L21:227 weakly suppresses
- L21:1886 weakly suppresses
- L21:764 weakly suppresses
- L21:1451 weakly suppresses
- L20:226 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
- Releases
- ht
- losing
- lists
- men
- 领导人
- signings
- roducing
- 列表
- Trials
Suppresses
- capacity
- .capacity
- Capacity
- 责任感
- ức
- capacity
- Sense
- _capacity
- ency
- 对公司