the word 'put'
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. 19 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 responds to the word 'put' and its forms, as in 'help you put', 'put together' and 'put a lid on'. Pushing it up did not produce clearly put-related text; the outputs wandered across topics.
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.
-
bone density test called a DEXA, which looks for low bone density as a possible cause of the injury. A team of sports medicine experts can help you put shin splints behind you and keep running strong.⏎—Dr. Jordan Metzl
-
confusion.⏎Dividing with Decimals⏎Many times students freeze at the sight of a decimal, but there’s no need to do so. Let me help you put the decimal drama to rest.⏎Can
-
experiences and the people around her. My advice to you is stop worrying about whether your teen’s attitude measures up to other teens’ attitude but to encourage her to put her best attitude forward at all times.⏎Why? Because her attitude affects her behaviour
-
Ready to start? Let’s dive in!⏎A little history of the SSL certificate⏎They say “to know your future you must know your past”, right? Put into a closer, business perspective, I’d rephrase that with “to better
-
spokesperson, Tim Farron, enjoys a strong base of support in the<|endoftext|>Many of us are inclined not to talk about things that upset us. We try to put a lid on our feelings and hope that saying nothing will be for the best.
-
, windy days are one of the worst. Having spent plenty of days battling a headwind on a ‘get out for a spin’ type of ride, we put together a few basic tips which we use at AATR to help us through
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: some of the the world is not well, and some of the other the world is not well. the world is not well for the first, the second, and the last of the 1999, their share of the
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
and immature Elizabeth Ann and follows her through a stunning metamorphosis into an independent and vibrant Betsy.⏎Fisher’s keen intelligence, delightful writing style, and perceptive nature shine through the pages of Understood Betsy. While the story was
-
tables, and not with any of the other⏎storage engines. Also, this warning against sharing a data⏎directory among servers always applies in an NFS environment.⏎Permitting multiple MySQL servers to access a common data⏎directory over NFS is a
-
powder that was invented by Hakaru Masumoto at Tohoku Imperial University in Sendai, Japan, about 1936 as an alternative to permalloy in inductor applications for
-
(more information)⏎©Copyright 1993 by the Regents of the University of California⏎AND IS MAINTAINED FOR ARCHIVAL PURPOSES ONLY⏎Perennial to trees, from membranous bulb, fibrous corm, sc
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
- L17:214 excites
- L0:498 excites
- L6:875 suppresses
- L0:497 excites
- L2:1554 excites
- L18:103 suppresses
- L1:6 suppresses
- L18:196 suppresses
- L7:300 excites
- L10:237 excites
- L15:381 excites
- L15:129 suppresses
- L18:1380 excites
- L7:2731 weakly suppresses
- L18:133 weakly suppresses
- L18:1104 weakly excites
- L17:350 weakly suppresses
- L17:517 weakly suppresses "but" in "not only X, but" phrases
- L17:241 weakly excites
- L16:260 weakly excites
- L2:3071 weakly excites
- L18:139 weakly suppresses
- L16:989 weakly excites
- L18:322 weakly excites
Feeds later layers
- L21:518 suppresses
- L23:1398 weakly suppresses
- L21:2040 weakly suppresses
- L20:411 weakly suppresses
- L22:357 weakly excites
- L20:1908 weakly suppresses
- L20:283 weakly suppresses
- L24:782 weakly excites
- L20:1090 weakly suppresses
- L26:1255 weakly suppresses
- L27:155 weakly suppresses
- L20:648 weakly excites
- L23:670 weakly suppresses
- L26:2917 weakly suppresses
- L27:2928 weakly excites
- L20:2763 weakly excites
- L21:1264 weakly suppresses
- L21:2968 weakly suppresses
- L21:1188 weakly suppresses
- L24:2092 weakly excites "self" before esteem or image
- L25:199 weakly suppresses
- L24:1473 weakly excites
- L23:1206 weakly suppresses
- L24:1240 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
- worm
- ал
- 拯
- piece
- marshal
- 层出
- /latest
- edition
- сли
- えて
Suppresses
- nails
- amer
- anced
- }}">{{$
- utable
- iciente
- cente
- ducation
- Specifications
- uated