'the' or 'of' after a role or title
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. 25 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 lights up on the small word 'the' (or 'of') right after a phrase that sets up a job, title or role, like 'is the Messenger', 'known as the' or 'Minister of'. Turning it up made the model's text drift into unrelated topics and repeat itself, so the effect on the writing was messy.
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.
-
them as false gods.⏎The second part of this declaration is that we believe that the Holy Prophet Muhammad, may peace and blessings of Allah be upon him, is the Messenger of Allah. Since he has come for the final, universal religion of Islam
-
.D. 697-707, and was one of the great poets of the early days of Japan; he is known as the rival of Akahito Yamabe (see next verse), and after death was
-
did reach Istanbul in 1918, Russia had made a separate peace with Germany and declared itself the enemy of all bourgeois states, France and Britain prominent among them.⏎What made you
-
great struggle, we got freedom on 15th August 1947. He played a vital role in our independence, so we called him as the father of the nation. All our Indians with great love and affection called him as
-
came to India in 1917 after acquiring a high degree of doctorate in economics from Britain.⏎According to his agreement, he worked as the Minister of Finance in the princely state of Baroda for some time. But he did
-
greater now than it was before that Revolution which sought to bring about Islamist rule across the Arab world with the Shia Ayatollah Khomeini attempting to act as the leader of worldwide Islam. Khomeini hoped that Shi'ites in Iraq would
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 the "streamery" (ery, streamery) streamery streamery streamery streamery streamery streamery streamery streamery streamery streamery streamery streamery streamery streamery streamery streamery streamery stream
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
, that all men are created equal, that they are endowed by their Creator with certain unalienable Rights, that among these are Life, Liberty and the pursuit of Happiness.⏎Locke didn’t stop there, however
-
, that all Men are created equal, that they are endowed by their Creator with certain unalienable Rights, that among these are Life, Liberty and the pursuit of Happiness." Government's place in our lives is to provide us the basic God given
-
the interplay of human emotions, thoughts, and actions by taking mythological metaphors. Kamayani has personalities like Manu, Ida and Śraddhā who are found in the Vedas. The great deluge described in
-
could have been avoided⏎Jefferson tried to explain as best he can the philosophical principles of the war and of the DOI⏎life, liberty, and the pursuit of happiness⏎what Jefferson said that every human being had
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
- L20:1391 excites
- L7:247 suppresses
- L24:8 weakly suppresses
- L22:589 weakly excites
- L10:315 weakly excites
- L18:967 weakly suppresses
- L2:562 weakly suppresses
- L19:1228 weakly excites
- L24:573 weakly excites
- L10:2738 weakly suppresses
- L10:1042 weakly suppresses
- L19:827 weakly suppresses
- L20:1089 weakly suppresses
- L7:1314 weakly suppresses
- L4:1380 weakly excites
- L21:241 weakly excites
- L24:75 weakly suppresses
- L20:1159 weakly excites
- L19:175 weakly excites
- L20:677 weakly excites
- L15:1274 weakly excites
- L16:1378 weakly suppresses
- L10:2370 weakly suppresses
- L10:934 weakly suppresses
Feeds later layers
- L27:1972 weakly suppresses
- L26:2334 weakly suppresses the 'de' syllable starting words
- L26:1717 weakly excites
- L27:2522 weakly excites
- L27:2446 weakly excites
- L26:2499 weakly suppresses
- L26:1655 weakly excites
- L26:2460 weakly excites
- L26:2257 weakly excites
- L26:901 weakly excites
- L26:1719 weakly excites
- L26:2836 weakly excites
- L26:760 weakly excites
- L27:1731 weakly excites
- L26:2187 weakly excites
- L26:1376 weakly suppresses
- L27:984 weakly suppresses
- L26:1763 weakly excites
- L26:2230 weakly excites
- L27:2751 weakly suppresses
- L27:3018 weakly suppresses
- L27:1959 weakly excites
- L27:2299 weakly suppresses
- L27:1384 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
- Sql
- сти
- DbType
- upper
- DataBase
- sql
- lower
- UserDao
- şı
- sql
Suppresses
- ery
- stream
- [string
- 一行
- mitter
- iral
- -stream
- Stream
- String
- iet