comma in a list of paired nouns
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 responds to a comma placed between two or three items in a list, such as 'storage, management' or 'letters, numbers'. Turning it up produced rambling text with no clear link to the trigger.
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.
-
capabilities, and why should your business consider using it?⏎Put simply, Terraform is powerful infrastructure as code (IaC) tool that allows you to create, update, and control versioning of cloud infrastructure. With Terraform, you can
-
, caricatures and sheet music document opera's dramatic 400 year history. From the evocative to the humorous, these images capture opera's artistic, social, economic and political influence, as well
-
中具有较大的作用,使用率高,但是泰国学生经常在学习和使用汉语助动词中产生各种的偏误。本文通过调查法、统计
-
file represents a sequence of characters (letters, numbers or special characters) that are stored on a personal computer, accessible only to
-
ocean...⏎and hydrologic sciences⏎Hydrology is the study of the movement, distribution, and quality of water on Earth and other planets, including the hydrologic
-
improved.⏎What is the Cloud?⏎The cloud, in the context of business operations, refers to a network of remote servers hosted on the internet that enable the storage, management and processing of data and applications. Unlike traditional on-premises infrastructure
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: what is the value of the problem? I am not sure if it's something that is worth considering, and I think it might be something that is worth considering. I need to know if it's something that is worth considering. Let's
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
several technical challenges; (1) creating a robust shell with adequate wall thickness, (2) how to integrated the speaker grille into the plaster form, and (3) creating a suitable mold for the plaster that could could be re-used and not
-
al factors: unbalanced nutrition, excessive exercise and/or trauma⏎Most foals will be born with some degree of limb deviation, mostly
-
protect animals there. He began to talk to people about animals’ rights, he sought support for his views among his influential friends, he used his money as well as his diplomacy, and eventually he got some animal protection laws<|endoftext|>Play is one of
-
with patients who have speech and/or swallowing problems. Someone who has suffered a stroke definitely needs speech therapy in order to help him feed himself independently without choking as well as to restore his speech function. The speech therapist also works in schools with children who
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
- L7:1223 suppresses "and" joining two decades or times
- L9:749 excites
- L8:1238 suppresses
- L8:70 excites 'and' between stacked superlatives
- L8:593 suppresses
- L9:1802 suppresses
- L8:1046 excites
- L9:458 suppresses
- L7:689 excites
- L6:620 suppresses "and" in lists of three or more
- L7:519 suppresses
- L6:427 suppresses
- L3:250 excites
- L8:479 excites
- L6:584 suppresses
- L8:986 suppresses
- L7:498 suppresses
- L5:1469 suppresses
- L9:212 excites
- L8:1848 suppresses
- L6:1706 suppresses
- L9:2726 excites
- L8:410 excites
- L8:131 suppresses
Feeds later layers
- L13:656 excites 'and' or comma between paired items
- L11:127 excites
- L12:959 excites
- L11:251 excites
- L16:300 excites commas inside lists of nouns
- L13:517 excites comma closing a list item
- L13:125 excites
- L13:1275 excites
- L14:672 excites
- L12:1212 excites
- L15:395 weakly excites
- L12:225 weakly excites
- L11:682 weakly excites gap before a second number
- L12:1756 weakly excites
- L11:1595 weakly excites
- L13:346 weakly suppresses
- L16:761 weakly excites superlatives like 'largest' and 'first'
- L17:1499 weakly excites
- L13:933 weakly excites
- L12:690 weakly excites
- L15:1916 weakly excites
- L17:1510 weakly excites
- L16:2234 weakly suppresses
- L16:1766 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
- ocate
- роли
- {name
- ={}
- ammers
- $update
- {$
- <Role
- [var
- {*
Suppresses
- try
- trag
- ispiel
- swallow
- try
- swallowed
- ynth
- swallowing
- it
- puts