'for' after an award or honour
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. 18 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 strongly on the word 'for' when it introduces the reason someone was honoured, rewarded or punished, like 'won the Nobel Peace Prize for' or 'an award for'. Turned up, the model fell into repeated 'for the community' phrases.
Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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.
-
, Denison Miller was reappointed from 1 June 1919 for a further seven years. He was knighted in 1920 for services to<|endoftext|>But what exactly is Diabetes? There are a lot of myths and
-
she is the author of Bulldozer: Demolition and Clearance of the Postwar Landscape, winner of the 2017 Lewis Mumford Prize for the best book in American planning history. She is particularly interested in post-World
-
website, Equipe Cousteau , and the educational and scientific websites of the International Polar Foundation. In February 2009 CAML received an award for 'overall exellence' from its parent program, the Census of Marine Life
-
family was strongly loyal to the Confederate cause. Her older sister Eugenia married a U.S. congressman from Alabama, and was herself imprisoned several times by Federal authorities for her Southern nationalism. A younger brother, Samuel, was a major in the Georgia
-
the past⏎Before he became president of Poland in 1990, Lech Walesa had already won the 1983 Nobel Peace Prize for his dangerous human rights fight in the face of a communist government. His work eventually
-
numerous awards, including being named best business start-up in the Institution of Chemical Engineers Global Awards 2020 and being highly commended at the same event for energy saving.⏎Dr Feng and his team also have a start-up and are working
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 idea that a person can be in a state of [fill in the blank] [1] ] ] is not just a temporary, like a bird can be in a state of [fill in the] ]
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
, you need to do it soon. Failure to do so is a felony punishable by up to five years in prison or a $250,000 fine.⏎And the system has not been updated since women were allowed to serve in
-
, but also that it can monitor,<|endoftext|>This material must not be used for commercial purposes, or in any hospital or medical facility. Failure to comply may result in legal action.⏎WHAT YOU NEED TO KNOW:⏎Diphtheria is a disease
-
take in the privacy of our own homes, making it particularly convenient in our busy day-to-day lives.⏎This post may contain affiliate links. I will earn a small commission, at no extra cost to you, if you make a purchase through an
-
oil spilling should be fined and the laws should be tightened. I also wish the world would always be in peace
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
- L13:214 excites
- L12:2776 suppresses
- L10:2857 excites
- L14:2332 suppresses
- L15:2769 excites
- L16:2910 excites
- L11:1279 suppresses
- L12:468 suppresses
- L9:2582 excites
- L9:2099 excites
- L9:2435 suppresses
- L13:2329 excites
- L6:2553 excites 'for' after praise words like 'known for'
- L6:12 excites
- L8:1588 weakly suppresses
- L15:1804 weakly excites
- L6:617 weakly excites
- L7:1731 weakly excites
- L14:2607 weakly excites
- L13:378 weakly excites
- L17:210 weakly suppresses
- L7:1558 weakly suppresses
- L12:2852 weakly excites
- L9:2823 weakly excites
Feeds later layers
- L21:1609 strongly excites
- L22:2351 excites
- L21:2230 weakly excites
- L19:1796 weakly excites
- L21:1435 weakly excites
- L20:330 weakly excites
- L21:2368 weakly excites
- L19:931 weakly suppresses
- L20:2367 weakly excites
- L19:2117 weakly suppresses
- L23:2460 weakly excites
- L19:3050 weakly excites
- L23:1038 weakly excites
- L19:747 weakly excites
- L24:1162 weakly excites
- L26:82 weakly suppresses
- L26:1064 weakly excites
- L21:2729 weakly excites
- L23:553 weakly suppresses
- L20:1853 weakly excites
- L26:2970 weakly excites
- L20:2725 weakly suppresses
- L20:1804 weakly suppresses
- L19:757 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
- ��
- 功劳
- pons
- 行为
- haber
- positor
- ButtonItem
- CLU
- dish
- adv
Suppresses
- 限り
- enda
- arat
- ilogue
- われ
- relude
- hope
- ounter
- '%'
- 感覺