'of' inside institution names
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. 10 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 'of' inside the names of schools, institutes and organisations, such as 'Higher School of Economics' or 'Harvard T.H. Chan School of Public Health'. Its negative side is weak and mixed. Turning it up did not produce text that clearly relates to institution names.
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.
-
980-4369⏎DOMINGUES, Beatriz Helena. The guarani jesuitic missions in the context of Enlightenment. História [online]. 2006, vol.25, n.1
-
the Internet: legal frames and technological implications’ organized by the Higher School of Economics on the 7th Internet Governance Forum (Baku, Azerbaijan,
-
ium comprises articles on legal and technological issues of Internet Governance in sphere of human rights prepared by the group of legal and technical scholars of information studies of the Higher School of Economics. This compound is devoted to the workshop on ‘Human Rights on
-
Harvard T.H. Chan School of Public Health, Sept. 2012, https://www.hsph
-
hn School of Medicine at Mount Sinai.<|endoftext|>Visualise different economic scenarios, how they might be
-
checkout.⏎Feng ZW, Cao HF, Zhou XP (1999) The effects of acid precipitation on the ecological environment and restoration. Press of Chinese Environmental Sciences, Beijing, pp 1–3⏎Lei HC,
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: What is it, and what is it, and what is it, and what is it, and what is it, and what is it, and what is it, and what is it, and what is it, and what is it
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
power of a successfully communicated thought, from one human mind to another, is one of the greatest forces we know. But like the tango, it takes two to communicate
-
ance Inc. illustrates how this can be applied to a major change initiative. The company’s CEO initiated a major organizational change to move from a more traditional, top-down culture to a servant-leadership culture. A 18
-
, it also means that pollutants that originate indoors are getting trapped inside.⏎“People normally<|endoftext|>Spring Dawn by Meng Haoran 春眠不觉晓 处处闻啼茑 夜来风雨声 花落知多少⏎Pin
-
aching a wider audience effectively! : Because an instructor can handle a small mob at a time only. When every other student bombards him with questions and raises hand in every micro second, the teacher sits down with head in his hands.⏎· It
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
- L4:42 suppresses
- L9:207 suppresses
- L4:92 suppresses
- L9:789 excites
- L9:46 suppresses
- L3:130 excites
- L3:23 suppresses
- L9:2581 excites
- L7:92 excites
- L3:672 excites
- L4:267 excites
- L3:249 weakly excites
- L8:277 weakly excites
- L7:2621 weakly suppresses
- L3:2650 weakly excites
- L9:1139 weakly suppresses
- L3:615 weakly suppresses
- L2:678 weakly suppresses
- L4:76 weakly excites
- L9:991 weakly suppresses
- L2:974 weakly excites
- L2:1592 weakly suppresses
- L9:1470 weakly suppresses
- L9:429 weakly excites
Feeds later layers
- L11:2472 weakly excites
- L25:2562 weakly suppresses
- L17:2512 weakly suppresses
- L23:542 weakly suppresses
- L27:131 weakly suppresses
- L24:1133 weakly suppresses
- L17:1404 weakly suppresses
- L15:1226 weakly excites
- L20:1726 weakly suppresses
- L25:574 weakly excites
- L19:1549 weakly suppresses
- L13:2397 weakly excites
- L25:170 weakly excites
- L15:2225 weakly excites
- L22:1403 weakly suppresses
- L25:2149 weakly suppresses
- L11:2147 weakly suppresses
- L24:777 weakly suppresses
- L23:1806 weakly suppresses
- L15:2795 weakly suppresses
- L13:825 weakly excites
- L16:2715 weakly suppresses
- L18:1833 weakly excites
- L17:2758 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
-
- 第一批
- ян
- gänge
- 责任人
- 每个
- 第一次
- 法定代表
- огда
- 第三个
Suppresses
- rij
- udo
- orial
- irement
- Signup
- Signup
- كم
- abra
- rement
- exas