'of' after a noun
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. 8 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' when it follows a noun, as in 'the results of', 'the formula of' or 'copyright of'. When it was turned up strongly, the output broke into Chinese text and filler characters, so the steering result does not match the trigger.
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.
-
If you haven't already taken a look at the articles, I hope you will find some time to check them out. I encourage you to read the first article of the following list.⏎If you want to liven up your chemistry course with some
-
signed up to receive the notifications of new studies, you can do that today (the form is at the bottom of
-
productivity of cast aluminum, are widely used in the production of semi-continuous casting and die castings.⏎Foam Ceramic Filter Manufacturing Process⏎The main raw material of the porous filter material is high-temperature alumina. It does not react with
-
. But, where registration is available, many authors register anyway, especially for works that are sold for money. That is because registration helps to prove that the copyright of a work belongs to a certain author.⏎If an author gets paid to make a
-
standard contribution is used when the marginal cost is using.⏎Sales Volume Variance Formula⏎In general, the formula of Sales Volume Variance is⏎Sales Volume Variance = (Actual units sold –
-
questions give students a sneak preview of what to expect on the test. Furthermore, students can look at the results of a practice test to determine which skills they need to work on.⏎Some students may
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 fact that, in the context of a non-profit, the development of a certain amount of information and which is directly related to the content of the information, the information will first and foremost be determined by the level and the overall situation of
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
unique way and this has created dualities of value that is often difficult for others to understand (Chattergee 1991, Garg/Parikh 1995). Other than any other countries Indian managers gave lot more
-
will protect your home against air pollutants<|endoftext|>Active and Passive Voice – Tips for Content Writing⏎February 22nd, 2012 by Donna Mathews⏎The use of active voice is important while writing online content. It improves the
-
and teachers?⏎According to the Montana Common Core Standards<|endoftext|>Copyright © 1995 Brother Eric Vogel, St. Mary's College⏎Balsamorhiza hookeri Nut
-
Effective technical communication is honest, clear, accurate, comprehensive, accessible, concise, professional in appearance, and correct.” – Mike Markel⏎Technical writing is used in many technical and professional fields. Its goal is to
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
- L1:25 suppresses
- L5:176 suppresses
- L7:92 weakly suppresses
- L3:53 weakly excites
- L5:136 weakly excites
- L5:2120 weakly excites
- L3:249 weakly excites
- L0:769 weakly suppresses
- L6:13 weakly excites
- L3:171 weakly suppresses
- L2:1570 weakly suppresses
- L3:805 weakly suppresses
- L4:342 weakly excites
- L7:1734 weakly suppresses
- L7:3013 weakly suppresses
- L4:471 weakly excites
- L3:2817 weakly suppresses
- L7:3034 weakly suppresses
- L2:2912 weakly excites
- L3:1870 weakly excites
- L4:552 weakly suppresses
- L3:2731 weakly suppresses
- L3:1172 weakly suppresses
- L1:1227 weakly suppresses
Feeds later layers
- L9:281 strongly excites
- L9:207 strongly excites
- L10:370 suppresses
- L9:46 suppresses
- L9:147 excites
- L18:477 suppresses
- L11:840 suppresses
- L10:1032 excites
- L9:537 excites
- L21:525 excites
- L9:232 excites
- L18:233 suppresses
- L17:758 excites
- L16:1519 excites
- L20:395 excites
- L17:2202 excites
- L10:602 excites
- L19:2338 excites
- L18:646 excites
- L18:950 weakly excites
- L10:197 weakly excites
- L9:530 weakly excites
- L26:687 weakly excites
- L15:3035 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
- Xiao
- jiang
- imizer
- итесь
- wang
- ้อน
- hua
- mak
- .ibatis
- тся
Suppresses
- CppType
- 也为
- 回合
- 腓
- �回
- Beth
- ölçü
- 数额
- uspendLayout
- Brussels