'as' after 'use' or 'taking' (as a model)
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. 17 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 'as' when it says something is used as an example, model, or standard, like 'use them as general guidelines'. Turning it up made the model produce repetitive, loose text, so steering doesn't clearly show this.
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.
-
STE as technical names and technical verbs). STE gives you the rules and the categories for these technical words. Basically, writers can use the approved words in the dictionary as a core vocabulary. But they can also use terms that are usual in their companies
-
where glycemic index comes into the scene. Glycemic index (GI) classifies foods based on how quickly they elevate blood sugar levels, taking pure glucose as the comparison point.⏎Foods with a high GI, such as white bread and
-
ous Explorations Through Science and Stories (MESS)⏎Teachers may use this guide and recommended materials as a resource to respond to children’s interest and curiosity as they explore the properties of
-
, read across to find your target heart rate. Your maximum heart rate is about 220 minus your age. The figures are averages, so use them as general guidelines.⏎|Age||Target HR Zone 50-85%
-
This Title The Easter Rising: Revolution and Irish Nationalism, Second Edtion⏎In this innovative work, Alan Ward uses the pivotal event in twentieth-century Irish history as a prism through which to survey Irish history from the twelfth century to
-
and science for every school year group. We really hope that you can use some, or all, of this content as you see fit, whether that’s to compliment, vary or boost work set by
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: you have to be with the code. It's not the code you're doing it on the computer, but you're there there. You have to be with the code. Let this be. Let this be, because you're there.
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
was literally found at your fingertips. Fingerprints and signatures used to be a trustworthy component of your acceptance or approval – especially in matters concerning finance and law. As the world changes, most of our financial matters have moved seamlessly online. Our digital identity
-
of Oregon and Washington. Let's hope Cascadians welcome them back as the agents of ecological restoration that they are.⏎Posted by Eric de Place | Permalink
-
. We are pleased to see that learning more about a different form of communication could help us to help patients.<|endoftext|>As the planet warms, China’s wheat crops will be threatened by more frequent epidemics
-
While some people and teens are natural planners, most are not. And they may find it challenging to figure out how to “become” a planner.⏎As the saying goes…⏎Without a plan, teens will:⏎- lose time trying to figure
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
- L16:1780 suppresses
- L10:417 suppresses
- L9:1877 suppresses
- L3:1233 excites
- L6:2593 suppresses
- L5:1395 excites
- L4:683 excites
- L10:132 weakly excites
- L5:1998 weakly excites
- L10:714 weakly suppresses
- L5:165 weakly suppresses
- L16:792 weakly excites "the" inside a prepositional phrase
- L15:1151 weakly suppresses
- L15:232 weakly suppresses
- L16:2847 weakly suppresses
- L6:2224 weakly excites
- L5:2881 weakly suppresses
- L3:936 weakly suppresses
- L0:2810 weakly excites
- L5:1016 weakly excites
- L9:319 weakly excites equals signs in maths and equations
- L3:470 weakly suppresses
- L15:7 weakly suppresses
- L8:410 weakly excites
Feeds later layers
- L19:1874 excites
- L20:1253 excites
- L23:916 excites
- L19:2656 excites
- L19:1815 excites
- L22:1204 excites
- L19:2531 excites
- L19:2511 excites
- L20:2244 excites
- L19:2638 excites
- L19:2450 excites
- L26:588 weakly suppresses
- L18:1036 weakly excites
- L19:2553 weakly excites
- L24:2939 weakly excites
- L22:2991 weakly excites
- L22:684 weakly suppresses
- L18:90 weakly excites
- L18:144 weakly suppresses
- L25:2270 weakly suppresses
- L18:2233 weakly suppresses
- L21:2805 weakly excites 'the' right before 'subject' or 'source'
- L19:1906 weakly excites
- L21:1851 weakly suppresses
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
- part
- “
- follows
- follow
- ges
- ус
- “So
- 候选人
- part
- ”
Suppresses
- itudes
- itude
- rium
- BufferData
- estion
- angkan
- latina
- 经济技术
- neighbouring
- OrNil