summary transition phrases like 'in short'
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. 6 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them 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 words that sum up or wrap up a point, such as 'in short', 'all in all' and 'overall'. Turning it up made the written text drift into unrelated, rambling content, so the steering output doesn't show a clear link to the trigger.
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.
-
climes has had to wrestle with rapid changes, demands for food, clothing, shelter and fuel, relative scarcity of all these and difficulty of securing them—in short, nearly every possible element in his surroundings which would compel him to get out and
-
narrator’s and characters’ voices and which is richly composed of idiom and colloquialism, exclamations, repetitions, and circumlocutions—in short, all the features of a mind instantly verbalizing. In this way, the
-
-lost ship. It is believed that the Endurance is located more than two miles (3.2 km) beneath the surface of the ocean.⏎All in all, the team will spend about 45 days in the Southern Ocean, most
-
, and its land… In short, every Indian nation is free to adopt its own laws and be ruled by them
-
F-22 is perhaps best known for its role in the Global Strike Task Force. Each one costs about $148 million.⏎Needless to say, the F-22 is kind of a big deal. Going head to head
-
|⏎|Required to build||libstdc++|⏎|Required to build||fortran|⏎
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 question of how to approach the problem of writing a well-structured and well-organized argument. I know that some people approach the problem of writing a well-structured and well-organized argument. In this case, I have been trying
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
, 1/4 turn anti-clockwise and then forward 1_______.⏎9. Can you give Pirate Pete directions to find Percy Parrot?⏎Great work, Shipmates! Now Pirate Pete has gathered his things, he is ready to
-
acycline resistance was maintained in spite of the declining concentration of tetracycline. (C) 2002 Elsevier Science B.V. All rights reserved.⏎|Publication status||Published - 2002|⏎- whole
-
of Scouts BSA.⏎Printable helps for requirements:⏎Get started on your Next Big Thing Nova award with these ideas and related achievements:⏎STEM stands for science, technology, engineering, and mathematics. The NO
-
When you have almost any queries concerning wherever along with tips on how to employ online yoga certification, you’ll be able to contact us with our own site. This word is derived the Sanskrit root, yog, which means to combine. Our ultimate
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
- L3:542 excites
- L5:1004 weakly excites
- L4:2058 weakly excites
- L5:1143 weakly suppresses
- L4:270 weakly suppresses
- L4:226 weakly excites
- L5:341 weakly excites
- L4:567 weakly suppresses
- L3:853 weakly suppresses
- L5:372 weakly suppresses
- L4:773 weakly suppresses
- L1:266 weakly excites
- L5:1171 weakly suppresses
- L3:2068 weakly excites
- L5:1550 weakly suppresses
- L4:543 weakly suppresses
- L4:29 weakly suppresses
- L5:1463 weakly excites
- L5:3035 weakly suppresses
- L5:1771 weakly excites
- L4:2515 weakly excites
- L4:2286 weakly suppresses
- L3:386 weakly excites
- L5:326 weakly suppresses
Feeds later layers
- L8:510 excites
- L9:705 excites
- L16:1687 suppresses
- L7:1253 excites
- L11:3070 weakly excites
- L18:2610 weakly excites
- L9:40 weakly suppresses
- L7:1714 weakly excites
- L9:1517 weakly suppresses
- L17:1029 weakly suppresses
- L13:1734 weakly excites sentence end right before a line break
- L7:796 weakly excites
- L9:2277 weakly suppresses
- L13:2735 weakly excites
- L7:181 weakly excites
- L9:2459 weakly suppresses
- L10:1716 weakly excites
- L8:169 weakly excites
- L8:960 weakly excites
- L13:240 weakly excites
- L14:2312 weakly excites
- L8:1839 weakly excites
- L22:2940 weakly excites
- L10:2380 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
- Nationals
- -elect
- -redux
- Examiner
- 锦标
- fusion
- Trot
- 给您
- | ⏎
- @(
Suppresses
- ilent
- 频
- 日消息
- akes
- ada
- UILD
- ools
- aç
- апр
- esti