the word 'sales'
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. 25 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 'sales' in business and marketing text. When it was turned up, the model started repeating the word 'call' over and over.
Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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.
-
. A white project manager with similar training and work experience was recently hired, and you find out that he will be paid more than you. You are a top salesperson for your
-
, including financial resources, market surveys, performance indicators and competitor performance statistics.⏎Business financial resources that may indicate strength during your SWOT analysis include financial resources derived from sales activities, income from investors and the value of company infrastructure. Strong financial performance will
-
on strengths. Performance indicators include rate of return on investment, production rates, sales growth, stock dividends, labor turnover rates and market share.⏎When you have completed
-
for the product.⏎You could make each category a separate view with each view in a block. Each view can have different criteria. You can give your CEO, sales manager, and marketing manager access to the block admin page and let them fight over
-
and oral);⏎- Marketing/selling skills (e.g., sales, publicity, fundraising);⏎- Scientific skills (e.g., lab skills, scientific
-
— is essential to any sales team. It enables teams to focus on exploring leads, fostering customer relationships, and
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: an economy that is physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical physical
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
of criteria by which we all agree that it is “done”. This is a more rigorous definition than in other agile methodologies, where it is variously described as a “promise of
-
prototype?⏎- What are the benefits of low fidelity prototypes?⏎- What do wireframes look like?⏎- What is fidelity prototype?⏎- What is the main disadvantage of high fidelity prototyping?⏎- What is wireframe mockup
-
to coding best practices such as modular and reusable code, efficient algorithms, and proper documentation enhances the app’s maintainability and scalability. Embracing agile development methodologies and collaboration tools promotes effective teamwork and streamlines the development process.⏎Furthermore, continuous testing
-
UML represents a compilation of "best engineering⏎practices" which have proven successful in modelling large, complex systems.⏎UML succeeds the concepts of Booch, OMT and OOSE by fusing them into a single,⏎common and widely
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
- L3:1558 suppresses
- L21:311 suppresses
- L18:779 suppresses
- L20:1140 weakly excites
- L0:2071 weakly excites
- L4:2629 weakly suppresses
- L6:2784 weakly suppresses
- L6:2007 weakly excites
- L9:2091 weakly excites
- L9:2356 weakly excites
- L22:2056 weakly suppresses
- L17:2308 weakly suppresses
- L21:1121 weakly excites
- L21:873 weakly suppresses
- L6:2944 weakly excites
- L9:2486 weakly suppresses
- L17:2956 weakly excites
- L5:1655 weakly suppresses
- L21:2944 weakly excites
- L8:1832 weakly suppresses
- L10:2740 weakly excites
- L16:2390 weakly suppresses
- L11:2301 weakly excites
- L10:1286 weakly suppresses
Feeds later layers
- L27:2652 weakly suppresses
- L26:2214 weakly excites
- L27:1437 weakly excites
- L26:2203 weakly excites
- L27:1012 weakly suppresses
- L26:1690 weakly suppresses
- L26:1917 weakly suppresses
- L27:418 weakly suppresses
- L27:1656 weakly excites
- L27:2835 weakly suppresses
- L26:2916 weakly suppresses
- L27:2048 weakly suppresses
- L27:2063 weakly excites
- L27:2185 weakly suppresses
- L26:1640 weakly suppresses
- L27:3008 weakly suppresses
- L27:29 weakly excites
- L27:2409 weakly excites
- L27:2189 weakly excites
- L26:1404 weakly suppresses
- L27:2698 weakly suppresses
- L27:2790 weakly suppresses
- L27:1375 weakly excites
- L26:1163 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
- call
- führt
- OLL
- calls
- olla
- call
- klä
- 冈
- ナイ
- oll
Suppresses
- raphic
- economy
- ted
- Economy
- imm
- aking
- ewan
- econom
- economies
- Econom