'and' or 'or' joining list items
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. 14 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 'and' or 'or' when it joins two nouns or short items in a pair, like 'food and drink' or 'ceramics and composites'. When it was turned up, the model's text drifted into lists and repeated numbering.
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.
-
experimentally derived films. Good agreement between model and experiment was observed.⏎ASJC Scopus subject areas⏎- Electronic, Optical and Magnetic Materials⏎- Ceramics and Composites⏎- Condensed Matter Physics⏎- Materials Chemistry<|endoftext|>Humans are intimately
-
⏎Stores other vitamins and minerals (fats, iron, copper, A, D, K, and
-
. In addition to vocabulary, there are also sentences and idiomatic expressions. New Amigos also includes geographical information and cards focused on culture, business, and food and drink.⏎New Amigos is a great game<|endoftext|>How did the kindly Christian saint
-
's Revised Unabridged Dictionary1. (n.) A thought or a consideration offered in support of a determination or an opinion; a just ground for a conclusion or an action; that which is offered or accepted as an explanation; the efficient cause
-
achieve an appropriate ‘fit’ with this environment. This includes:⏎Results that meet the expectations of its owners (shareholders, government, members etc. )⏎Products and services that meet its clients’ expectations at least as well as rivals’⏎Ability
-
direct link between eating nourishing foods and higher grades, alertness, and better memory and recall.⏎One simple reason for this is that ensuring you consume protein, carbs,
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 if we are trying to create a more sustainable future, and we want to build a better world, and we want to bring the attention to the importance of this, that we are to be our own leaders, but we want to have
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
une-system⏎- Kennedy, David O. “B Vitamins and the Brain: Mechanisms, Dose and Efficacy—A Review.” Nutrients, vol. 8, no. 2, Jan. 201
-
. (2021). Contrasting response of native and non‐native plants to disturbance and herbivory in mountain environments. Journal of Biogeography, 48(7), 1594-1605
-
to finance its assets. The WACC is the minimum acceptable return that a company must earn on an existing asset base to satisfy its creditors, owners, and other providers of capital, or they will invest elsewhere.⏎The following is the
-
Shale Revolution⏎- The Health Impacts of the Shale Revolution⏎- The Effects of the Shale Revolution on Local Governments⏎- The Shale Revolution and Water Quality⏎Common Resources — Feb 7, 2018⏎
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
- L6:798 excites
- L9:92 suppresses "and" or "or" joining two items
- L13:977 suppresses
- L13:131 excites 'and' joining parallel phrases
- L4:57 suppresses
- L11:406 suppresses
- L10:335 excites commas in catalog, date, and list metadata
- L6:427 suppresses
- L9:77 excites commas in citations and name lists
- L5:880 excites
- L4:3 excites commas after citation fields
- L7:2041 suppresses 'or' offering an alternative
- L7:836 excites
- L9:2010 suppresses
- L9:1942 excites
- L10:74 weakly excites
- L10:554 weakly excites
- L8:70 weakly excites 'and' between stacked superlatives
- L9:2244 weakly excites
- L12:1786 weakly excites
- L12:1235 weakly suppresses commas inside numbers and lists
- L4:2270 weakly excites
- L6:912 weakly suppresses
- L7:1223 weakly suppresses "and" joining two decades or times
Feeds later layers
- L21:735 excites 'or' in ranges and alternatives
- L15:1501 excites "and" joining two parts of a name or phrase
- L17:572 excites
- L15:1162 excites
- L17:2274 excites
- L15:2831 excites
- L17:1214 excites
- L15:734 excites
- L17:2386 excites
- L18:2136 excites
- L18:870 excites
- L16:1179 excites
- L15:2722 excites "and" or "or" in a pair or list
- L18:1270 excites
- L19:883 excites
- L16:340 excites
- L16:2580 excites 'and' linking food or health nouns
- L16:300 excites commas inside lists of nouns
- L16:1774 excites
- L18:1386 excites unclear: words before 'and' or 'or'
- L16:1369 suppresses
- L18:1625 excites
- L19:1823 excites 'and' inside organisation or paired names
- L16:1074 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
- /or
- phans
- equival
- isons
- consist
- acles
- combinations
- nbsp
- even
- дет
Suppresses
- “(
- Stories
- орт
- ariant
- amework
- ughter
- elic
- etration
- Award
- ascular