page and volume numbers in citations
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. 13 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 spot right before page or volume numbers in bibliography entries, such as after "pp." or after a journal volume. The negative side fires inside names and words, like the apostrophe in a spacecraft name. Turning it up made the model write repetitive or filler text that does not clearly show this pattern.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
"Purchasing Power Parity Exchange Rates for the Global Poor,"⏎American Economic Journal: Applied Economics,⏎American Economic Association, vol. 3(2), pages 137-66, April.⏎- Angus Deaton & Olivier Dup
-
"Purification Study of Pervious Concrete Pavbement," International Journal of Engineering and Technology vol. 5, no. 5, pp. 532-535, 2013.<|endoftext|>Following
-
1722. This text is based on Kessinger Publishing's facsimile (ISBN 0-7661-6894-8) of J. and P. Knapton's edition of 1
-
in Education (eJSiE):⏎2, Article 1.⏎Available at: http://epublications.bond.edu.au/ejsie/vol8/iss2/1<|endoftext|>Some may argue that sleeping is one of the best parts
-
Contrasting response of native and non‐native plants to disturbance and herbivory in mountain environments. Journal of Biogeography, 48(7), 1594-1605 https://doi.org/10
-
the board exams. Students can easily NCERT Solutions for Class 10 Maths Exercise 6.5 Chapter 6 Triangles from below.⏎CBSE Class 10 Maths Chapter 6 Triangles Exercise 6.5 Solutions⏎Topics
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 told that I have been given a question and I have been told that I have been given a question that is worth 1000, and I have been given a question that is worth 1000,
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
moon that we earthlings don’t see and has yet to be studied—that is, until now.⏎On January 3rd, China’s spacecraft called the Chang’e 4 made a soft landing on the dark side of the moon. Its mission
-
-reaching consequences.⏎In addition to being as a major milestone in the history of astronomy and the study of Deep Sky Objects. One of these objects is known as Messier 10 (aka. NGC 6254), a glob
-
, 2, 3, 4, 5] i = len(given_arr) - 1 while(i >= 0): print(given_arr[i]) i = i - 1⏎In this program,⏎- given_arr
-
tapered edge. The positive side, or anode, typically has no marking.⏎Decode the diode type and material designation. This is designated by the "1N" in the part number. The "1" identifies the component as a di
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
- L9:884 excites
- L8:2591 excites
- L12:752 excites
- L10:574 excites
- L6:371 suppresses
- L9:1108 suppresses
- L10:2461 excites
- L12:1301 excites
- L12:1809 suppresses
- L12:2110 suppresses
- L11:1218 suppresses
- L12:713 excites
- L0:2404 excites
- L6:1469 weakly excites
- L12:2956 weakly suppresses
- L4:973 weakly suppresses
- L10:281 weakly excites
- L11:293 weakly suppresses
- L12:1319 weakly suppresses
- L12:2168 weakly excites
- L12:1496 weakly excites
- L11:2835 weakly excites
- L8:350 weakly suppresses blank space before a year
- L5:2869 weakly excites
Feeds later layers
- L15:958 suppresses
- L17:1744 suppresses
- L14:1542 excites
- L14:677 excites
- L17:1806 suppresses
- L15:1041 excites
- L17:163 suppresses
- L16:2099 suppresses
- L17:202 excites
- L16:2059 suppresses
- L15:42 suppresses
- L14:1094 suppresses
- L17:354 suppresses
- L18:92 suppresses
- L16:302 suppresses
- L18:352 suppresses
- L15:748 weakly suppresses
- L17:932 weakly suppresses
- L14:193 weakly excites
- L16:2148 weakly suppresses
- L15:289 weakly excites
- L17:222 weakly suppresses
- L14:371 weakly excites
- L14:453 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
- ertext
- rots
- possible
- (Set
- арт
- ETS
- SAT
- ropriate
- PT
- elman
Suppresses
- 发生了
- 决定了
- HasKey
- 发生
- 凡是
- washer
- oplay
- 發生
- ()}}⏎
- 涵盖了