'with' after a verb of action
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. 6 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 reacts to the word 'with' when it comes right after an action verb, as in 'covered it with' or 'responded to ... with'. Turning it up did not produce a clear change in the model's replies.
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.
-
deck (it can be the teacher, or the players can take turns picking a card), and if a player has that picture on their board, they cover it with a penny or a dried bean. The winner is the
-
Antwerp, the Belgian athlete Omar Smet, who was a member of the KAA Gent athletic department, responded to the rhythmical cheering of the American delegation with the student yell ‘Buffalo! Buffalo!’ Together with him, KAA Gent
-
arguing that the Brazilian Constitution defines a "family entity" as "a stable union between a man and a woman."2 The Catholic Church responded to the recent ruling with outrage. As Archbishop Anuar Battisti put it, the Supreme Court's decision
-
taller growth. Too much direct sunlight can scorch the leaves. Put them 3 to 4 feet away from a bright window, or shade the window lightly with a sheer curtain.⏎Keep the room temperature between 65 and 80
-
eagle shrieks and leopard growls), the researchers found greater complexity in expression than previously understood as well as differences in alarm calls between the two locations.⏎Confirming with linguistic means some hypotheses initially made by primatologists, their analysis showed that these
-
How many times like this are there between midnight⏎and 7 a.m.?⏎These practical challenges are all about making a 'tray' and covering it with paper.⏎My cousin was 24 years old on Friday April 5th
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 not a fan of this. I am not a fan of this. I am not a fan of this. I am not a fan of this. I am not a fan of this. I am not a fan of this.
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Stage - A guide for parents'. This outlines information for each phase of development aligned with each area of learning and also shares ideas on how you can support your child at home. Here are some key pages that may be of particular use:⏎Pages
-
making appointments with them and any typical front desk protocols.⏎- Support your child in using their voice and express any concerns they have to their doctor.⏎- Keep a
-
attractiveness to the 21st Century mind.⏎At the turn of the 20th Century raged a debate (begun in the 1870s in America) in legal circles in the US
-
across multiple pages. The pattern of rhyming two lines and ending with a dark, dark cave continues. This sets a mood. The reader is set up for something a bit spooky or mysterious in only 14 words.⏎In all three
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
- L2:2762 excites
- L3:1837 weakly excites
- L5:2547 weakly excites
- L3:1095 weakly excites
- L3:122 weakly excites
- L1:1961 weakly suppresses
- L5:1021 weakly suppresses
- L5:2401 weakly suppresses
- L4:2442 weakly suppresses
- L5:1794 weakly excites
- L3:115 weakly suppresses
- L5:1164 weakly excites unclear: weak, word-piece and 'Note' labels
- L2:2824 weakly excites
- L0:483 weakly excites
- L4:2452 weakly excites
- L3:2441 weakly suppresses
- L3:2174 weakly suppresses
- L5:1401 weakly suppresses
- L1:1960 weakly excites
- L3:1385 weakly excites
- L4:1845 weakly suppresses
- L5:1254 weakly excites
- L5:659 weakly excites
- L3:63 weakly excites
Feeds later layers
- L9:1340 excites
- L13:1237 excites
- L17:2667 excites
- L14:1953 excites
- L15:1836 excites
- L13:831 excites
- L14:2374 excites
- L14:2022 excites
- L12:2685 suppresses
- L15:527 suppresses
- L15:2443 excites
- L7:428 excites
- L17:2355 excites
- L20:1525 excites
- L8:984 suppresses
- L10:2488 weakly excites
- L17:1721 weakly excites 'by' introducing a method or means
- L23:2763 weakly excites
- L11:2015 weakly excites
- L16:1717 weakly excites
- L8:390 weakly excites "a" after sentence-opening "As"
- L12:719 weakly excites
- L17:1839 weakly excites
- L8:1901 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
- FontAwesome
- 什麽
- 阿拉伯
- mere
- parenthesis
- MBER
- トー
- )}.
- казал
- COME
Suppresses
- noqa
- .TODO
- antaged
- pliant
- tractor
- �能
- TRL
- icult
- ports
- ritic