'the' or 'as' before a superlative role
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. 9 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 small words like 'the' or 'as' that come right before a phrase describing a special role, such as 'the clear leader', 'the most recycled' or 'emerges as an important'. It seems to be getting ready for a word like 'most', 'only' or 'leader'. When it was turned up, the model's replies drifted into unrelated topics, so the steering did not show a clear theme.
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.
-
get there, the company plans on conducting an orbital test of the spaceship later this year.⏎By nabbing the Artemis contract, SpaceX has cemented itself as the clear leader in the private spaceflight world. One could argue that it already did
-
than producing steel from ore.⏎In 2011 households across Europe saved the equivalent of almost half Belgium’s annual electricity consumption by recycling packaging.⏎Steel remains the most recycled
-
of the lot. It is also the least spoiled today, because 34,718 acres of it above
-
yield, plants generally show a reduction in the rates of photosynthesis and growth, often associated with changes in plant metabolism. Nitric oxide (NO) has emerged as an important signaling molecule in
-
nicely integrate the existing materials with the two-dimensional materials," said lead researcher Yeonwoong Jung, an assistant professor at UCF's NanoScience Technology Center and the Materials Science & Engineering
-
sea or land.⏎With its large population, financial and legal services, cultural and historical attractions, and “Old Port” dining and shopping district, the City is the only true metropolitan area in the state. It is the county seat of Cumberland County
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 I have been thinking about the world around me, the way I have been thinking about the people who have been thinking about the people who have been thinking about the people who have been thinking about the people who have been thinking about the
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
ate, write the correct conjugate acid or base for the species in the pair.⏎a. H2O, OH-⏎Frequently Asked Questions⏎What scientific concept do you need to know in order to solve this
-
that all the questions answers in concern with this chapter are prepared by our expert teachers team. Here you will find three Pdf files, among them first Pdf files consists of Textbook Questions Answers whereas in second Pdf files you will get the additional extra questions
-
term Olive Ash does not refer to any specific species of Ash (Fraxinus genus), but instead is in reference to the darker, streaked heartwood found in some Ash trees, which
-
division is not enough to explain what happens to the different types of carbohydrates in the body.⏎For example, starches in the bread and potatoes are classified as complex carbohydrates. However, the body converts starches to blood sugar as rapidly as it processes
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
- L7:724 excites
- L7:2432 excites
- L5:668 suppresses
- L7:1437 weakly suppresses
- L0:2398 weakly excites
- L2:1880 weakly suppresses
- L5:641 weakly suppresses
- L6:2224 weakly excites
- L8:769 weakly excites
- L7:1054 weakly suppresses
- L5:613 weakly suppresses
- L1:2513 weakly excites
- L0:2107 weakly excites
- L5:505 weakly suppresses
- L4:656 weakly suppresses
- L1:1766 weakly suppresses
- L7:3064 weakly excites
- L6:2509 weakly excites
- L3:2598 weakly excites
- L1:1318 weakly excites
- L5:1317 weakly suppresses 'a' right after 'being'
- L7:1016 weakly excites 'something' after is or not
- L7:1265 weakly suppresses
- L0:1871 weakly suppresses
Feeds later layers
- L17:1013 excites
- L23:916 excites
- L20:1253 excites
- L18:1036 excites
- L14:827 excites open bracket after a value or citation
- L10:417 excites
- L14:1402 excites
- L19:654 excites
- L14:783 weakly suppresses
- L16:1711 weakly excites
- L22:1204 weakly excites
- L16:2038 weakly excites
- L11:217 weakly excites
- L17:1959 weakly excites
- L17:1245 weakly suppresses
- L16:282 weakly excites
- L27:1368 weakly excites
- L17:2698 weakly suppresses
- L16:2743 weakly excites
- L14:549 weakly suppresses blank spot right before a number
- L19:2553 weakly excites
- L24:546 weakly excites
- L26:490 weakly excites the digit 1 starting a number
- L17:2252 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
- 一团
- /**<
- 唯一的
- UPPORT
- @brief
- REQUIRED
- @section
- 唯一
- 一回事
- ))"⏎
Suppresses
- фи
- rases
- bef
- fame
- 等活动
- arrang
- форм
- né
- Reputation
- iped