'etc' ending a list
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. 17 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 word 'etc' when it closes a list of examples, like 'caries, etc' or 'fishing, etc'. Turning it up made the model fall into repeated 'like that' phrases, which does not clearly connect to lists.
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.
-
develop a definitive diagnosis, which allows the dental professional to rule out other possible causes of the pain (periodontal disease, caries, etc) and then implement an appropriate treatment plan to address the problem.⏎Dentin is
-
quick to blame low deer numbers due to drought, coyotes, poor habitat, etc., the plain and simple fact is that deer had an abundant food source available to
-
and lining the page, figuring out text length, divisions, etc.⏎Writing technology has made it easier for us to focus more on the text and
-
their race, religion, sex, or some other factor that is specified in the law. Unless there is a crime like assault, aggravated assault, attempted murder, etc. hate-crime laws generally do not apply.⏎Hate speech legislation criminalizes
-
onslaught of a second wave.⏎Major hotel chains, businesses focusing on labor-intensive work, service providers, etc. are among the major categories. It would include the employers of these categories in
-
extreme heat, entrained air, moisture, contamination by dirt and debris, inadvertent mixing with different oil, etc. - that degrade the integrity of the hydrocarbon base stock and deplete the
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: this is that this way like this this this this this this this this this this this this this this this this this this this like this this this this this this this this this like this this this this this this this like this this this this
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
and well-being. Most people think being a dentist means putting your fingers in people’s mouths and looking for cavities, but there’s a lot more to it than that. Dentists work in lots of places outside the dental office, including hospitals
-
fields,<|endoftext|>Bronze Age Odin⏎Since many want to be a leader it is easy to see Odin as the leader of the tribe. I doubt it always was so. Others maybe saw it as the leading aspect in man. In the Ritual
-
pressure for nervous writers.⏎Setup for Revealing Moment Exercise⏎First have students simply draw a line on a piece of paper. (It doesn't get any simpler than this). You may want to do this on the board along with them.⏎Then
-
, periodontitis, and possibly even obesity.⏎"People thought that asthma was caused by inhalation of allergens but this study shows that it may be more complicated than that - asthma may involve colonization of the airways by multiple bacteria," said study
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
- L15:1930 suppresses
- L15:1237 suppresses
- L4:1352 suppresses
- L16:2861 excites
- L15:1995 weakly suppresses
- L16:942 weakly excites
- L14:1316 weakly excites
- L15:2683 weakly suppresses
- L16:666 weakly excites
- L6:1376 weakly excites
- L0:229 weakly excites
- L16:1991 weakly suppresses
- L6:2654 weakly suppresses
- L7:493 weakly suppresses 'this' pointing back to a topic
- L12:3027 weakly excites 'other' in 'and other ...' lists
- L14:1814 weakly suppresses
- L15:398 weakly excites
- L10:1603 weakly suppresses
- L6:773 weakly suppresses
- L10:2724 weakly excites
- L6:1425 weakly excites
- L16:2420 weakly excites
- L16:2692 weakly excites
- L16:1999 weakly excites
Feeds later layers
- L20:1922 suppresses
- L20:1704 suppresses
- L18:2938 suppresses
- L23:2015 suppresses
- L22:1469 suppresses
- L19:2918 suppresses
- L18:2923 weakly suppresses
- L21:492 weakly suppresses
- L18:313 weakly suppresses
- L18:769 weakly suppresses
- L20:42 weakly excites
- L25:769 weakly suppresses
- L23:932 weakly excites
- L21:1376 weakly suppresses
- L26:405 weakly suppresses
- L19:1875 weakly suppresses
- L19:1265 weakly suppresses
- L18:686 weakly excites
- L27:694 weakly suppresses
- L20:1766 weakly suppresses
- L20:2949 weakly suppresses
- L19:1124 weakly suppresses
- L19:1646 weakly suppresses
- L19:2425 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
- бир
- '>"+
- obe
- yers
- opoly
- orgh
- 经验和
- ibName
- еб
- aversable
Suppresses
- 如此
- così
- 这样
- 然
- así
- such
- 那样
- thus
- 这样的话
- Such