'Well' starting a reply or answer
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. 1 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 'Well' when it starts a sentence that answers a question, like a conversational reply. Turning it up made the model's writing drift into random topics, so the effect on style is unclear.
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.
-
leave the helmet at home? Well…it depends.⏎Should Snowboarders Wear Helmets?⏎It has been argued that
-
V. That’s 10% of the intended voltage!⏎So what are the traditional options? Well, the best one is to have a separate set of wires known as “sense
-
some old style logic that needs to be reviewed.⏎Why not just set the maximum length possible? Well…⏎A VARCHAR is subject to the character set it’s in, for UTF8
-
stuff on the pool floor? Well you might actually have an outbreak of mustard algae, also known as yellow algae.
-
you remember when you learned multiplication and discovered multiplication was a faster and shorter way to represent repeated addition? Well there is a faster and shorter way to represent repeated multiplication.⏎Fractions - Adding
-
, what about the link between a woman’s menstrual cycle and the lunar cycle? I mean, that’s why we’re all still reading this post, right?⏎Well, let’s start off with a bit of backstory: For generations, a woman
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: I have been working on this project in the last week, and I need to take care of you, but I don't think I can do that. I need to take care of you, but I don't think I can do that
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
10 0 10 why am I getting 10 0 20.I am confused can some one help me in this regard.thanks in advance. kamesh. here is the code from Marcus Mock exam
-
the most common types of tomatoes available. They can be meaty like a beefsteak, but always juicy.⏎Vesuvio Piennolo<|endoftext|>Thank you to DadsWorksheets.com for providing a VARIETY of math practice sheets
-
medical advice and should not take the place of a doctor’s visit. Please see the bottom of the page for more information or visit our Terms and Conditions.⏎Thanks for your feedback.⏎37 of 42 found this helpful<|endoftext|>Television was
-
Subject: Overt Markers on Mass Nouns⏎E-mail this message to a friend⏎Query for this summary posted in LINGUIST Issue:⏎Thanks for the many responses on the topic of count/mass marking on nouns.⏎The issue
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
- L0:178 suppresses
- L0:1301 suppresses
- L0:1224 suppresses
- L0:2520 weakly suppresses
- L0:419 weakly suppresses
- L0:2789 weakly suppresses
- L0:30 weakly excites
- L0:1870 weakly excites
- L0:2085 weakly excites
- L0:1663 weakly suppresses
- L0:2203 weakly suppresses
- L0:2222 weakly excites
- L0:2433 weakly suppresses
- L0:31 weakly suppresses
- L0:2744 weakly excites
- L0:1436 weakly excites
- L0:2794 weakly suppresses
- L0:1266 weakly suppresses
- L0:2747 weakly suppresses
- L0:2004 weakly excites
- L0:447 weakly excites
- L0:268 weakly excites
- L0:1003 weakly suppresses
- L0:1869 weakly excites
Feeds later layers
- L8:2883 weakly excites
- L15:1602 weakly suppresses
- L13:668 weakly excites
- L27:1427 weakly excites
- L20:2835 weakly excites
- L14:2803 weakly suppresses
- L2:1688 weakly suppresses 'off of' and 'of' after 'off'
- L9:1925 weakly suppresses
- L4:1964 weakly excites
- L8:2227 weakly suppresses
- L8:2277 weakly suppresses
- L6:1530 weakly excites
- L15:2064 weakly suppresses
- L5:2543 weakly suppresses
- L17:1689 weakly excites
- L18:1944 weakly suppresses
- L15:1502 weakly excites
- L13:2075 weakly excites
- L14:2533 weakly suppresses
- L11:1191 weakly excites
- L9:520 weakly excites
- L22:2103 weakly excites
- L13:250 weakly suppresses
- L10:2315 weakly suppresses line break inside wrapped text
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
- lates
- iances
- trận
- "](
- тан
- ][_
- iors
- 暮ら
- iores
- STDOUT
Suppresses
- him
- fall
- ARRY
- 观点
- 6
- sem
- 观念
- consum
- server
- Mp