word before him or he
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. 26 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. 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 a word that is usually followed by a pronoun such as "him", like "led", "tells", "earned" or "enables" in "led him away". It is slightly negative on opening quote marks. Turning it down or up changed the text in unrelated ways, so steering does not confirm the pattern.
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.
-
ing Spiritualism and Materialism’, Ludwig Andreas Feuerbach wrote- “Der Mensch ist, was er ißt.” (Man is what he eats) and by this,
-
can choose his own life. Man has dignity; he is intrinsically and individually valuable. He owns himself and the products of his labor. His rational nature enables him to build a free society where he is not a slave to his rulers.
-
most senior position.⏎Fisher wrote to Miller on 14 May 1912 offering him the position of Governor. Miller accepted and his Commission appointing him to the role was signed on 29 May 1912
-
in Venice, where his father was in practice. He was destined for the law, but his inherited tastes led him away from jurisprudence to medicine. He practised in several different towns
-
.⏎When Hamlet is engaged in a discussion with Gertrude, they are sparring with words. She is talking about one thing, he another. She tells him he has offended his father very much (meaning Claudius, who is not
-
feelings with his northern neighbor. He nevertheless won a major victory over the Edomites. Afterward, however, his veneration of the gods of Edom earned him a prophetic condemnation.⏎Seeking to punish Israel for raids against Judean
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: It sounds illogical that you could get you you you you you you you you you you you you you you you you you you you you you you you you you you you you you you you you you you you you you you you you
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
social scientist, and cognitive psychologist whose primary research interest was decision-making within organizations and is best known for the theory of “bounded rationality” and the theory of “satisficing.” He applied psychology to statistics, operations, and decision-making scenarios
-
a violation of the most trivial rules.4 Another characteristic of this crushing form of economic Darwinism is that it thrives on a kind of social amnesia that erases critical thought, historical analysis, and any understanding of broader systemic relations. In
-
the Oregon National Primate Research Center at Oregon Health & Science University, monitored the reproductive systems of healthy female nonhuman primates following exposure to Delta-9-tetrahydrocannabinol, or THC, the main psychoactive ingredient in marijuana
-
Genetic algorithms used to be one of the cornerstones of what is called computational intelligence. Computational intelligence is a subfield of artificial intelligence based mainly on the idea of creating intelligence by mimicking natural processes. For example, swarms display a certain kind
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
- L24:477 strongly excites
- L24:2385 strongly excites
- L23:656 strongly suppresses 'the' in historical names
- L23:1052 suppresses
- L23:646 excites
- L24:440 suppresses
- L25:310 excites
- L23:3026 suppresses
- L21:1552 suppresses
- L23:346 suppresses
- L16:1238 suppresses
- L23:956 excites
- L22:520 excites
- L20:2754 weakly suppresses
- L25:2922 weakly suppresses
- L25:1971 weakly suppresses
- L22:2253 weakly excites
- L16:31 weakly suppresses
- L20:1446 weakly excites
- L16:39 weakly suppresses
- L16:1347 weakly suppresses
- L24:942 weakly suppresses
- L16:2464 weakly excites
- L23:2536 weakly suppresses
Feeds later layers
- L27:578 strongly suppresses
- L27:1244 strongly suppresses
- L27:737 weakly excites
- L27:717 weakly suppresses
- L27:486 weakly excites
- L27:1068 weakly excites
- L27:2412 weakly suppresses
- L27:1155 weakly suppresses just before a pronoun for God
- L27:544 weakly excites
- L27:1099 weakly excites
- L27:2190 weakly suppresses
- L27:1276 weakly excites
- L27:2641 weakly excites
- L27:2472 weakly excites
- L27:106 weakly excites
- L27:2972 weakly excites
- L27:3015 weakly excites 'F' starting a word in a split
- L27:2101 weakly excites
- L27:1399 weakly excites
- L27:2701 weakly excites
- L27:2135 weakly excites
- L27:1504 weakly suppresses
- L27:990 weakly suppresses
- L27:760 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
- orate
- viders
- rowsing
- 姐妹
- IFT
- orsk
- ее
- 坞
- eway
- 妇科
Suppresses
- he
- him
- 他
- 他在
- He
- 他也
- 他对
- 他说
- 他又
- he