scope words like all, many, suitable
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. 14 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 fires on words that describe how much or how widely something applies, such as 'all', 'many', 'across' and 'suitable'. Turning it up made the model repeat phrases about time and things without a clear point.
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.
-
which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.<|endoftext|>- Major Systems >⏎- The Plumbing Rough-In⏎The Plumbing Rough
-
lots of secrets.⏎- Because others are either sleeping or already have mates.⏎- Basic instincts.⏎- To get into the lime-light.⏎- In pursuit of a suitable alliance.⏎
-
on health expenditure and quality of care, building on the OECD expertise in these areas. Each indicator is presented in a user-friendly format, consisting of charts illustrating variations across countries and over time, a brief descriptive analysis highlighting the major findings conveyed by the
-
up to 10 minutes with each feeding...One bed bug will usually take more than one bite. Once a bed bug inserts its mouthparts and finds a suitable blood vessel, it will begin feeding. ~ source⏎While bed bugs like to
-
more than others.⏎Baits need to be specific to the catch you’re trying to make. Not all baits can be used for all types of fish. To put things into perspective, catfish love raw chicken liver
-
- Lessons align with National Health Education standards⏎- Lessons align with SEL programming and competency areas⏎Purchase grants one user, individual use of course materials for a self-selected period from date of purchase. This contract is extendable. Multi-year discounts available
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 story of the human race, as it exists in the heart of the novel. The story of the human existence on Earth, as a person. " " " " " " " " " " " " " " " " "
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
enhance public safety, and provides lifelong learning and technical exchange opportunities benefiting the global engineering and technology community. ASME has more than 127,000 members worldwide.⏎ASME′S VISION, MISSION STATEMENT AND CORE
-
.⏎Largest Cities in West Virginia⏎West Virginia is one of the few US states not to have a city with more than 100,000 residents. In fact, the largest city in the state, Charleston (the state
-
6, our organization in Chicago was called upon to assemble some detailed cost figures from butter manufacturers whose 77 plants produced almost 150,000,000 pounds of butter annually. These figures reflected accurately the big increases
-
and an astute political observer who, while she lived, played a key role in advancing her husband’s career.⏎Like us on Facebook and get articles directly in your news feed<|endoftext|>To Colored Men!⏎This poster was used to recruit recently freed
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
- L5:1072 suppresses
- L10:167 suppresses
- L12:2802 excites
- L12:2841 excites
- L12:1531 excites
- L13:595 weakly excites
- L11:2630 weakly suppresses weak 'for' and 'all' quantifier signal, unclear
- L6:826 weakly excites
- L13:851 weakly suppresses
- L9:278 weakly suppresses
- L11:1174 weakly suppresses
- L7:716 weakly suppresses
- L3:2571 weakly excites
- L1:629 weakly excites 'across' meaning spread over an area
- L4:918 weakly excites
- L4:1843 weakly excites
- L10:1697 weakly excites
- L12:1809 weakly excites
- L7:2758 weakly excites
- L11:918 weakly suppresses
- L4:2902 weakly excites
- L10:1683 weakly excites
- L10:875 weakly suppresses
- L9:1824 weakly excites
Feeds later layers
- L19:1854 excites
- L19:1833 excites
- L17:2388 excites
- L21:1486 excites
- L15:1019 excites
- L18:77 excites
- L16:1009 excites
- L19:2132 excites
- L17:1979 excites
- L19:1668 excites
- L15:1507 excites
- L16:982 weakly excites
- L15:428 weakly excites
- L15:2850 weakly excites
- L22:529 weakly excites
- L15:1840 weakly excites
- L16:1122 weakly excites
- L16:2556 weakly excites
- L18:2465 weakly excites
- L27:1170 weakly suppresses
- L15:2672 weakly excites
- L18:2058 weakly excites
- L18:309 weakly excites
- L21:808 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
- taxing
- Inflater
- ief
- 阵阵
- égor
- afi
- isky
- syn
- aise
- ỹ
Suppresses
- icles
- وص
- culos
- itations
- culo
- straints
- cord
- �자
- писан
- ulas