'question' after 'this'
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. 0 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 'question' when it follows 'this', as in 'this question'. It is a narrow pattern. Turning it up produced repetitive, unrelated text, so the steering shows nothing about 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.
-
this question by reviewing the scientific literature on cow-calf separation to provide perspective and, hopefully
-
generalization.)⏎Your reader will be thankful if you start with “This question…” or “This wig…” or, most likely, “This intrusion…” If
-
and less justifiable answers. Something like “Which of the following motivations best explains so-and-so’s actions in such-and-such novel?” Instead of posing this question, hoping that all your students take a moment to think about it, then hearing
-
2021, Fox News report, Tucker Carlson asked the question no one is really allowed to ask: “How many Americans have died after taking
-
? This question has now been answered thanks to Read more…⏎Tag: ancient inscription⏎12
-
motion and enterprise.⏎This Elibron Classics publication is a facsimile reprint of a 1901 variation by means of the Macmill<|endoftext|>This question seems like such a common occurrence. It most always follows the realization that we have
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: I am going, and I and the? I am...? I am going to...? I am going to...? I am going to...? I am going to...? I am going to...? I am going to...
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
the Cuban government of Batista. Subsequently, he became an important figure in Cuban diplomacy and a vocal critic of the United States and the Soviet Union. Later on he helped revolutionary groups in Congo and Bolivia until he was captured and executed by the
-
except one guy. The guy is named Bao Shuya, who was the one saved his ass and took him to Ju as a political asylum seeker and later on helped him regain his lordship. Duke’s displeased and asked why he’s
-
of feudal lords in China in the Eastern state of Qi, took shelter in the state of Ju when he was persecuted in a bloody coup by his uncle. Later on another coup took place over Qi’s court and he got to
-
story celebrates the important and impressive feat of a small but mighty creature.⏎Christy Ottaviano Books.⏎Hardcover Tiny Bird download PocketBook on Kobo.
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
None: this is the first layer.
Feeds later layers
- L5:2036 weakly excites
- L6:2668 weakly suppresses
- L4:446 weakly suppresses
- L5:202 weakly excites
- L17:1784 weakly suppresses
- L19:2031 weakly excites line breaks in web footers and metadata
- L14:1256 weakly excites
- L2:2040 weakly suppresses
- L9:1727 weakly suppresses
- L3:2440 weakly suppresses
- L13:42 weakly suppresses
- L7:2582 weakly suppresses
- L12:1260 weakly suppresses
- L21:161 weakly excites
- L2:17 weakly suppresses
- L19:412 weakly excites
- L5:1715 weakly excites
- L12:1542 weakly suppresses
- L1:1710 weakly suppresses
- L15:839 weakly suppresses
- L9:1352 weakly suppresses
- L17:1168 weakly excites
- L14:67 weakly excites
- L14:2834 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
- depend
- depends
- ependency
- 蝶
- 幻
- extrad
- -fe
- pest
- cosmetic
- tail
Suppresses
- 留言板
- том
- 录制
- acular
- ocalypse
- INCLUDED
- orial
- idges
- missão
- 服务能力