indefinite article before a noun
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. 10 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 words 'a' or 'an' that come right before a noun phrase, such as 'from a common ancestor' or 'on a small mountain'. It seems to be getting ready for the noun that follows. Turning it up did not produce a clear theme in the written output.
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.
-
of those things if evolution were true."⏎Austin wondered about the evolutionists' claim that human and ape DNA are so alike, it proves man and ape evolved from a common ancestor.and thus proves evolution.⏎Lisle pointed out their DNA does have many
-
URE as we should, we need to understand first of all that culture is a superstructure⏎thatrests upon a material basis.⏎The ideas, institutions and all cultural patterns are dependent on the material
-
Supreme Court’s decision inBrown v. Board of Educationis widely considered a seminal point in the battle to end segregation, but it was in fact the culmination of a decades-long legal campaign.Root and Branchis the epic story of the two fiercely dedicated
-
a default in lieu.⏎When the 406 Not Acceptable error message is received, it is in response to an HTTP request that includes a negation-related header such as Accept or Accept-Language.
-
o. Phnom Da in Takeo is one of the popular tourist attractions in Takeo.⏎Phnom Da, Takeo is a temple which was constructed on a small mountain. It was built in the 6th century by the King Rut
-
.⏎© Science Museum / Science & Society Picture Library<|endoftext|>Note: This lesson was originally published on an older version of The Learning Network; the link to the related Times article will take
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 will be in the country" in the park, and I will be there. And I will be there. And I will be there. And I will be there. And I will be there. And I will be there.
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
and learning is your purpose, you might have to invest a little to get valuable, extensive worksheets that actually assist your child to learn.⏎Worksheets that have been developed by a preschool instructor that knows what the school demands are and also what your youngster
-
fully accepted the highly abstract nature of Heisenberg's quantum mechanics, says Miller. They agreed with Galileo's assertion that "the book of nature is written in mathematics", but they also realized the power of using visual imagery to represent mathematical symbols
-
volcanoes, tsunamis, global climate, and weather.⏎Based on your research, do the following:⏎Support your statements with examples. Use a minimum of six reliable references, two of which should be peer-reviewed articles.⏎Write a
-
Christo Luksatrio, to encourage his program to educate students on 360 technology.⏎Luksatrio, who works as a technology development assistant for the School of Professional Communication, said he wanted to experiment with this type of
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
- L5:225 suppresses
- L9:656 excites
- L6:426 excites
- L8:191 suppresses
- L9:1081 excites
- L4:49 excites
- L7:298 suppresses
- L8:619 suppresses
- L9:457 excites
- L7:1880 suppresses
- L7:155 weakly suppresses
- L8:1997 weakly excites
- L8:193 weakly suppresses
- L8:294 weakly suppresses
- L9:942 weakly excites
- L2:1062 weakly excites
- L7:959 weakly excites
- L9:1724 weakly suppresses
- L9:1084 weakly suppresses
- L9:1811 weakly suppresses
- L8:1172 weakly excites
- L9:1228 weakly suppresses
- L7:1326 weakly suppresses
- L8:1000 weakly excites names of agencies and institutions
Feeds later layers
- L15:2513 excites
- L14:359 suppresses
- L12:966 excites
- L11:1257 suppresses
- L11:1960 excites
- L11:679 excites
- L18:717 suppresses
- L15:834 excites
- L11:1367 excites
- L12:1255 excites
- L14:1168 weakly excites
- L11:2336 weakly excites
- L18:1740 weakly suppresses
- L11:2087 weakly excites
- L13:2074 weakly excites
- L12:442 weakly suppresses
- L15:826 weakly suppresses
- L16:1178 weakly excites
- L13:1949 weakly excites
- L12:2041 weakly excites
- L16:2340 weakly excites
- L16:2643 weakly excites
- L19:1022 weakly suppresses
- L17:1542 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
- 又要
- setContent
- addCriterion
- oire
- 另一方面
- .setCursor
- 又是
- (PyObject
- ubishi
- 的对象
Suppresses
- 当
- bid
- lyph
- 当他
- thic
- rown
- even
- Atom
- Im
- bart