'yet' ending a negated clause
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. 2 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 'yet' at the end of sentences that say something has not happened so far, like 'haven't been tested yet'. Turning it up did not make the model's writing visibly about 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.
-
.⏎- Design & Thinking Official Trailer (02:28)⏎Design & Thinking is a documentary released this year about design thinking. I haven't seen it yet, but I'm loving the approach -- the cacophony of voices
-
became one of the first to author the word aight in published text. He’s even glossed it “(Alright)” for everyone who hadn’t caught on yet.⏎We learn from the OED entry that aight represents a colloquial
-
you boring and dull, because you have nothing to talk about, because everything you talk about is a memory. A person cannot talk about something that hasn’t happened yet in reality.⏎“Memory depends very much on the perspicuity, regularity
-
isn’t even a “hypothesis”.⏎Being a falsifiable hypothesis doesn’t mean the hypothesis is true...not yet. The hypothesis hasn’t been tested yet.⏎The hypothesis is never true or false by default. Only the tests or evid
-
gingivitis?⏎Gingivitis is the first and only stage that periodontal disease can be reversed. This is because the bones haven’t been attacked yet. Bleeding gums is usually one of the first symptoms experienced if you have ging
-
ifference-between-a-cote-de-and-pres-de⏎I hope this is helpful.⏎Bonne journée !⏎Sign in to submit your answer⏎Don't have an account yet? Join today⏎Test your French to the CEFR standard<|endoftext|>I believe that
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 role of a teacher in the field of Education. I have been thinking about lately: the role of a teacher in the field of Education. I have been thinking about lately: the role of a teacher in the field of Education. I
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
/restaurant companies who don’t use fair trade beans, for example). Socially responsible investing (SRI) is basically the practice of putting your money to work only in companies that are trying to make a positive impact on society.⏎You can pick
-
the item is re-sold at a higher price, then the profit will go down. With what is variable CPV, you don’t pay for the item only once but you get to sell it more times and therefore gain profit each time.⏎
-
.<|endoftext|>Our research goal is to understand the maximum short-term capabilities of electrical machines and investigate control methods to achieve them. Electric machine capabilities and limitations cannot be understood only through electrical specifications and nameplate data, because redundancy is included in the design.
-
are unique to each server. Otherwise, the servers try to log to the same files.⏎Even when the preceding precautions are observed, this kind of⏎setup works only with⏎MERGE tables, and not with any of the other⏎storage engines
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
- L1:2675 weakly suppresses
- L0:1011 weakly suppresses
- L1:583 weakly excites
- L1:1779 weakly excites
- L0:1825 weakly suppresses
- L0:2171 weakly suppresses
- L0:207 weakly suppresses
- L1:2445 weakly excites
- L1:1645 weakly excites
- L1:1356 weakly suppresses
- L1:2446 weakly suppresses
- L1:359 weakly excites
- L0:2634 weakly suppresses
- L0:649 weakly suppresses
- L0:816 weakly excites
- L1:2978 weakly suppresses
- L1:174 weakly excites
- L1:466 weakly suppresses
- L1:2830 weakly excites
- L1:316 weakly excites
- L1:445 weakly excites document start marker token
- L1:966 weakly suppresses
- L1:2642 weakly excites
- L1:2607 weakly suppresses
Feeds later layers
- L3:40 weakly excites
- L6:101 weakly suppresses
- L3:4 weakly suppresses
- L16:1477 weakly suppresses
- L27:425 weakly suppresses
- L13:2120 weakly excites
- L9:69 weakly suppresses
- L21:46 weakly suppresses
- L15:740 weakly excites
- L7:551 weakly excites
- L4:3058 weakly excites
- L26:2514 weakly suppresses the word 'just' after have or has
- L15:2751 weakly excites
- L23:234 weakly excites
- L3:1489 weakly suppresses
- L5:401 weakly suppresses
- L14:282 weakly suppresses
- L13:1034 weakly excites
- L21:1513 weakly excites
- L8:3001 weakly suppresses
- L13:1372 weakly suppresses number before an age or week count
- L4:2469 weakly excites
- L9:2622 weakly excites
- L8:149 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
- 략
- nc
- قات
- 赀
- 盛会
- craft
- mise
- �
- 坊
- ty
Suppresses
- "}>⏎
- igham
- ighbor
- olleyError
- <Student
- <footer
- nEnter
- /org
- /MIT
- ApiController