"to" in "referred to"
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 "to" in phrases like "referred to as" or "often referred to", where a term is being named or explained. Turning it up made the writing repetitive and oddly circular, which does not clearly relate to this phrase.
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.
-
60 and Goguryeo in 668. Thereafter, Unified Silla or Later Silla, as it is often referred to, occupied most of the Korean Peninsula, while the northern part re-emerged as
-
of DNA by combining existing nucleotides with new bases, creating new types of DNA molecules called DNA-based polymers. The DNA-based polymers, often referred to as designer genes, are used in a variety of ways to produce new proteins,
-
America with a mass market media campaign through the<|endoftext|>KAA Gent is often referred to by its second name: the Buffalos. At every match, you will
-
is?⏎You might have come across this term “cobots” while searching for a solution to improve your production processes but weren’t quite sure what the term referred to. As you’ve probably noticed yourself, cobots are trending more and more in
-
being.⏎What is blood cholesterol? Also referred to simply as cholesterol, blood cholesterol refers to how much of a lipid (fat)
-
are often referred to as early stage and late stage.⏎Many asthma sufferers who inhale an allergen
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: the way to the way of thinking, the way of thinking, the way of thinking... so I have been thinking about lately... the way of thinking... the way of thinking... the way of thinking... the way of thinking... the way
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
matter, such as compost. It thrives in full sun but will tolerate part shade.⏎To get the highest yield of tender and flavorful leaves, pinch the tip of each branch, starting in early summer when the plants are 6" tall,
-
conomic and moral rights⏎- Guarantee control over the work.⏎- Guarantee remuneration for work through selling or licensing.⏎- They are transferable.⏎- Right of attribution - the right to claim the authorship.⏎- Right of integrity - the
-
, & entrepreneurship all in one contest⏎Receive a FREE Maker-Kit full of inventing supplies from Picoturbine of $200 value (Limit of 3 per school) & a $200 Stipend for your
-
complete this third step by drawing the legs and the arms as shown in the illustration below.⏎On top of the head, draw antennas using broken lines. The tip of both antennas is made from a small
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
- L0:1224 weakly suppresses
- L0:178 weakly suppresses
- L0:778 weakly excites
- L0:24 weakly excites
- L0:1301 weakly suppresses
- L0:2738 weakly suppresses
- L0:516 weakly excites
- L0:2030 weakly suppresses
- L0:774 weakly suppresses
- L0:1987 weakly suppresses
- L0:486 weakly suppresses
- L0:664 weakly excites
- L0:2209 weakly suppresses
- L0:139 weakly excites
- L0:2579 weakly suppresses
- L0:150 weakly excites
- L0:1839 weakly excites
- L0:785 weakly suppresses
- L0:781 weakly excites
- L0:2818 weakly suppresses
- L0:2375 weakly excites
- L0:1481 weakly excites
- L0:60 weakly excites
- L0:78 weakly excites the 'S' in 'U.S.'
Feeds later layers
- L16:1088 suppresses 'of' after a noun phrase, as in 'of course'
- L15:785 suppresses 'of' in set phrases like 'of course'
- L19:715 suppresses
- L17:1835 suppresses
- L13:132 weakly suppresses "of" after a noun
- L14:1519 weakly suppresses
- L17:317 weakly suppresses 'of' after a noun phrase
- L4:4 weakly suppresses
- L17:385 weakly suppresses
- L7:790 weakly suppresses
- L3:315 weakly suppresses 'of' after a noun phrase
- L5:1207 weakly suppresses 'of' after abstract nouns
- L7:2458 weakly suppresses
- L6:371 weakly excites
- L22:229 weakly suppresses "of" after a noun in "X of Y"
- L17:1177 weakly suppresses
- L16:2194 weakly suppresses
- L2:506 weakly excites
- L15:2562 weakly excites
- L9:2264 weakly suppresses 'of' after superlatives and 'the X of'
- L3:712 weakly suppresses unclear: 'of' after a noun
- L17:2234 weakly suppresses
- L2:1274 weakly suppresses
- L6:1121 weakly suppresses 'while' joining two clauses after a comma
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
- izedName
- 秕
- aved
- 被打
- 又被
- mojom
- ning
- 会发生
- ифика
- 主义者
Suppresses
- erto
- odal
- lio
- erw
- 行列
- satisfaction
- 资产负债
- olia
- erta
- śli