words right after an 'if' condition
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. 11 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. 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 lights up on the word just after a conditional 'if' clause in advice, such as 'if you can't find', 'if you are unsure' or 'if you have trouble'. Turning it up made the model drift toward maths questions and 'what is the value' style prompts, which has little to do with the trigger.
Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens an email to the maker with the neuron already filled in).
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.
-
and chapter.⏎If you think you should have access to this title, please contact your librarian.⏎To troubleshoot, please check our FAQs, and if you can't find the answer there, please contact us.<|endoftext|>Indigenous Peoples and Climate Change
-
a pencil or marker to sketch your newspaper layout onto your blank paper.⏎- Search through magazines and newspapers for useful shapes, images and corresponding words. If you can’t find the perfect words, create them by cutting out individual letters.⏎- Experiment with
-
...⏎however you want Santa to look!).⏎- Close template window when done printing to return to this screen.⏎- Set page margins to zero if you have trouble fitting the template on one page (FILE, PAGE SETUP or FILE, PRINTER
-
whether you need further assessment or testing for COVID-19.⏎If you are unsure whether you need medical care, call HealthLink BC at 8-1-
-
PDF [portable document format] printable page. (If you are unable to view the printable page, download the free Adobe Acrobat Reader.)⏎Click for the
-
miss a dose of Ampicillin and are taking it regularly, take it as soon as possible. If it is almost time for your next dose, skip the missed dose and go back to your regular dosing schedule. Do not take 2
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 first" "the last" "the first" "the last" "the first" "the last" "the first" "the last" "the first" "the last" "the first" "the last"
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
or wanting to hear. A word to describe the process of removing one’s self (and one’s own responses) is the term “mindfulness”. When we become mindful of our own biases, tendencies, and prejudices, then we are better able
-
there are a slew of possible reasons for girls’ lack of access to opportunity, one worth considering is the opportunity cost. If education brings little else but personal development, it’s easy to see why few families make it a priority.⏎But what if
-
coaching, mentoring and supporting.⏎Those teachers who are at the peak of the profession are our best resource in delivering on our most ambitious goals for young Australians. If all young people are to achieve success in a worthwhile education, teachers must demonstrate careful planning
-
share that they failed to see the logic of working to the same basic voltage in selecting their architecture, so as to be more compatible with one another. If they were compatible
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
- L10:750 suppresses
- L7:1616 weakly suppresses
- L10:2062 weakly suppresses
- L10:534 weakly suppresses
- L6:2323 weakly suppresses
- L10:2088 weakly excites
- L4:2769 weakly suppresses
- L9:1347 weakly excites
- L10:2934 weakly suppresses
- L9:356 weakly excites
- L8:1866 weakly suppresses
- L7:1571 weakly suppresses
- L7:292 weakly suppresses
- L1:1131 weakly suppresses
- L8:691 weakly excites
- L6:1453 weakly excites
- L6:162 weakly excites
- L7:1880 weakly excites
- L1:1790 weakly excites
- L9:291 weakly suppresses
- L8:2468 weakly excites
- L10:1940 weakly suppresses
- L5:2588 weakly excites
- L3:2783 weakly suppresses
Feeds later layers
- L13:2785 excites
- L14:2642 suppresses
- L14:1481 suppresses
- L13:2364 excites
- L13:171 excites
- L12:268 excites
- L13:1273 excites
- L12:381 excites
- L13:1434 excites
- L12:2310 excites
- L14:2517 excites
- L12:193 weakly excites
- L12:2374 weakly excites
- L14:1709 weakly excites
- L14:1284 weakly suppresses
- L13:2223 weakly excites
- L14:2978 weakly excites
- L14:1126 weakly excites
- L18:1876 weakly suppresses
- L16:1342 weakly excites
- L13:2582 weakly suppresses
- L16:2228 weakly excites
- L16:931 weakly suppresses
- L14:2211 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
- genden
- jours
- 𝜂
- opard
- ocities
- orage
- contrat
- ляр
- ieux
- ouns
Suppresses
- 熙
- 温室
- psychology
- xbc
- rí
- attributes
- ICC
- _attributes
- dictionaries
- ares