'as' in 'as well as' and 'as much as'
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. 17 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 "as" in set phrases like "as well as", "as much as possible" and "as a result". Turning it up did not produce any clear change in the text the model wrote.
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.
-
Oracle With Clause Syntax and Examples :⏎In this section i would like to explain the syntax as well as examples of With clause in oracle.Before checking the syntax and examples of With
-
protect the earth’s climate. The global climate crisis is non-negotiable. #Move the Date: We need to do everything we can to delay the date as much as possible.⏎At VAUDE, we have already set out to reduce
-
With Clause is useful in Recursive queries as well.⏎These are some important points of With Clause in oracle. In following section i would like to explain about the syntax as well as multiple examples of with clause.⏎With SQL_Query_Name As⏎Select *
-
(University of Manchester), whose expertise lays in sustainable foods and food groups.⏎The event attracted an audience of 70, including the members of the general public as well as staff and students from the University of Nottingham and featured a lively debate on
-
work on the same piece? This section needs some elaboration. --EvanProdromou⏎Collaborative graphic design using software as a mediation device has some disadvantages, compared to collaborative development of text.⏎- Little
-
topic, which will take some time. World War 2 is such a broad theme that in order to write a profound essay, you need to narrow your topic as much as possible. Otherwise, you will write about everything and nothing at the same
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: 2-12-2-12-12-3-3-2-13-3-3-3-24-3-28-2-3-2-4-3-2
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
ecosystem approach principles into policy instruments, planning processes, and sectoral plans (e.g., in forest, fisheries, agriculture). At a more local level the principles of the ecosystem approach can be used to guide the development
-
schemes are rarely outweighed by the socioeconomic benefits accrued to translocated farmers," they write. "At an average start-up resettlement cost of at least US$12,00
-
and carried high-quality weapons and armour.⏎A key part of a Carolingian soldier’s armour was the shield which was critically essential for use on the battlefield at a time when lances
-
harm, similar to viruses. The average computer user arguably has no need to understand the intricacies of spyware, or any other form of malware. But at a minimum, they should be cognizant that it is malicious software that collects personal
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
- L6:328 strongly suppresses
- L2:8 strongly suppresses
- L10:103 strongly excites
- L0:481 strongly suppresses
- L2:1699 strongly excites
- L1:542 strongly suppresses
- L7:520 strongly suppresses
- L16:84 excites
- L2:1607 suppresses
- L5:138 suppresses
- L11:689 excites
- L1:20 suppresses
- L7:1581 suppresses the last zero in large numbers
- L16:75 excites
- L16:628 suppresses
- L5:779 suppresses
- L1:262 suppresses
- L3:1169 suppresses
- L1:13 excites
- L15:2537 suppresses
- L0:49 excites
- L7:2095 suppresses
- L1:1607 excites
- L2:1912 suppresses
Feeds later layers
- L18:102 strongly excites
- L20:174 strongly excites
- L19:569 strongly excites
- L20:121 strongly suppresses
- L18:205 suppresses
- L20:413 excites
- L20:183 suppresses
- L18:474 suppresses
- L19:645 suppresses
- L22:2749 excites
- L19:551 suppresses
- L18:198 suppresses
- L18:687 suppresses
- L23:2938 excites
- L21:1072 excites
- L18:750 suppresses
- L18:249 suppresses
- L18:936 suppresses
- L18:139 suppresses
- L21:921 suppresses
- L18:365 suppresses
- L18:328 suppresses
- L18:405 suppresses
- L19:390 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
- holes
- cents
- hole
- AndGet
- pues
- speaker
- lapse
- plane
- hyper
- synthesize
Suppresses
- ользоват
- clud
- erv
- Habit
- 皇家
- ernel
- channelId
- 자를
- èmes
- ecimal