'be' after 'to' or 'need to'
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. 20 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 lights up on the word 'be' when it comes after 'to' or 'need to', as in 'have time to be' or 'expects to be ready'. Turning it up made the model write in a loose, repetitive way, so the steering results don't clearly show what it does.
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.
-
⏎Why Lesson Planet?⏎Find quality lesson planning resources, fast!⏎Share & remix collections to collaborate.⏎Organize your curriculum with collections. Easy!⏎Have time to be more creative & energetic with your students!⏎Students are introduced to the characteristics of a
-
⏎Why Lesson Planet?⏎Find quality lesson planning resources, fast!⏎Share & remix collections to collaborate.⏎Organize your curriculum with collections. Easy!⏎Have time to be more creative & energetic with your students!⏎Students explore the Orange Revolution. In this
-
Why Lesson Planet?⏎Find quality lesson planning resources, fast!⏎Share & remix collections to collaborate.⏎Organize your curriculum with collections. Easy!⏎Have time to be more creative & energetic with your students!⏎Students participate in a variety of physical activities
-
<|endoftext|>Why Lesson Planet?⏎Find quality lesson planning resources, fast!⏎Share & remix collections to collaborate.⏎Organize your curriculum with collections. Easy!⏎Have time to be more creative & energetic with your students!⏎In this verb tense review worksheet, students
-
a Dragon spacecraft from the Cape Canaveral Air Force Station. The successful flight demonstrated the components necessary to transport humans to low-Earth orbit. SpaceX expects to be ready to fly its first manned mission in 2014.⏎Dragon is
-
-10. The full text can be found on the Board of Studies website.⏎Students must study at least 100 hours in one language, to be completed over one continuous 12-month period. When students have completed the mandatory
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: I do it, it means it does it, it does it, it means it does it. I mean it does it it *does* it it *does* it it *does* it. The it works it *does*
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
food and shelter. Fruits such as acorns or berries are vitally important, but sometimes other benefits are less noticeable. Many animals are host dependent--that is, they need a specific type of plant to survive. Butterfly larvae may chew on
-
. Recent research by Dr. Joseph Allen, a professor at the University of Virginia, points out the benefits of certain types of arguing between parents and children. That is, some forms of arguing actually strengthen
-
the Sensee River. Some of the men escaped the hail of bullets, but few⏎of the horses escaped. Tender-hearted gunners took time out to<|endoftext|>Is PET Plastic Safe for Food and Beverage?⏎Is PET plastic safe?
-
your personal desktop machine or laptop, and to files located on a remote web hosting server. We know how important it is to keep all the data safe. That is why at NTC Hosting we have developed a powerful backup solution, which takes care
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
- L18:960 strongly suppresses
- L17:227 strongly suppresses
- L19:1115 excites
- L19:576 suppresses
- L4:20 suppresses
- L15:599 excites
- L9:66 excites
- L16:466 excites 'is' after a long subject phrase
- L4:2128 suppresses
- L19:1612 excites
- L3:1301 excites
- L10:1736 suppresses
- L16:2038 suppresses
- L0:207 excites
- L18:612 weakly suppresses
- L17:2353 weakly excites
- L17:2050 weakly excites the word 'be' after modals
- L18:1167 weakly suppresses
- L6:2237 weakly excites
- L4:117 weakly excites
- L13:775 weakly suppresses
- L18:144 weakly suppresses
- L13:612 weakly suppresses
- L18:391 weakly excites
Feeds later layers
- L22:550 strongly excites
- L27:1368 excites
- L23:360 suppresses
- L21:1555 excites
- L22:2991 excites
- L23:576 excites
- L23:1160 excites
- L21:1814 excites
- L22:415 suppresses
- L21:1872 excites
- L21:1434 excites
- L26:395 excites
- L23:473 excites
- L23:1293 excites
- L21:1047 excites
- L23:1083 excites
- L23:1004 excites
- L26:571 excites
- L27:217 excites
- L25:2151 excites
- L23:628 excites
- L21:1675 excites
- L26:237 excites
- L23:1012 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
- considered
- flown
- able
- ready
- taken
- cie
- gone
- worth
- absorbed
- supposed
Suppresses
- edList
- must
- 日消息
- befind
- ющую
- 眼镜
- rop
- drop
- очередь
- 一点