unclear: personal names, mostly Indian
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. 4 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 tends to fire on parts of people's names, especially Indian and Hindu names such as Ram, Surendra and Shyam. The pattern is not very strong, and the steering results do not show a clear theme.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
etc. There is no confusion in the first name. The middle name can be seen from example – Ram Prasad Srivastava. Here First Name is Ram, Middle name is Prasad and Last name is Srivastava.⏎What
-
he toured South India extensively and composed songs on the various deities of the religious pantheon. He had met both Tyāgarāja and Śyām<|endoftext|>Different cultures have been using psychedelic substances in rituals for thousands of years. In
-
18).⏎Air Transport Law: Air transport law legislation aims safety of air passengers and implements rules regarding carriage. And also, ensures the passengers' rights to redressal for personal injury or damage.⏎2.1<|endoftext|>Vitamin D has at
-
of freedom and it is freed from the clutches of British rule. During the partition of Bengal and freedom by the indigenous movement in which Sir Henry Cotton Surendra Nath Banerjee and turned into a
-
birth, deciding its:⏎- ↑ Yogasutras 2.73⏎- The Concise Encyclopedia of Hinduism, Swami Harshananda, Ram Krishna Math, Bangalore<|endoftext|>Lancashire food producers and farmers have been working with
-
with Kasturba makanji. After that, he went to London to complete his bar studies. Later he worked at Indian firm in South Africa. While living in South Africa protest a war against racism in South Africa.⏎In 1
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 problem is a triangle with the given side. The problem is a triangle with the given side. The triangle is formed by the two sides. The triangle is formed by the given side. The triangle is formed by the given side. The
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
all are waiting for. Teachers, parents and student, now you will find below the Solutions of Class 10 Science Chapter 4 Environmental management questions answers. From the subject English of Maharashtra State Board.⏎Below you can see
-
class 10 appearing for the board exams. Students can easily NCERT Solutions for Class 10 Maths Exercise 6.5 Chapter 6 Triangles from below.⏎CBSE Class 10 Maths Chapter 6 Triangles Exercise
-
⏎(iv) For series RLC circuit at resonance, the phase angle is zero and cos . ϕ =1⏎∴ Pav = VRMS IRMS⏎Consider an AC circuit in which there is a phase angle of ϕ between VRMS
-
2012 June 23⏎Explanation: As seen from Frösön island in northern Sweden the Sun did set a day after the summer
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
- L3:62 excites
- L3:318 suppresses
- L3:249 excites
- L3:615 suppresses
- L3:506 weakly suppresses
- L3:153 weakly suppresses
- L1:2908 weakly suppresses
- L3:691 weakly excites
- L0:1768 weakly suppresses
- L3:421 weakly suppresses
- L3:530 weakly suppresses
- L3:2369 weakly suppresses
- L1:2278 weakly excites
- L2:1578 weakly excites
- L3:1907 weakly excites
- L3:189 weakly excites
- L2:2010 weakly suppresses
- L0:1515 weakly suppresses
- L2:2282 weakly excites
- L1:2677 weakly suppresses
- L1:2502 weakly suppresses
- L1:3007 weakly excites
- L3:880 weakly suppresses
- L0:1539 weakly suppresses
Feeds later layers
- L8:369 weakly suppresses 'of' after 'learned' or 'because'
- L5:2705 weakly excites
- L8:423 weakly excites
- L5:667 weakly excites
- L10:2186 weakly suppresses
- L25:2901 weakly suppresses
- L14:1515 weakly suppresses
- L19:967 weakly excites
- L10:2752 weakly excites
- L5:2599 weakly excites
- L6:1695 weakly excites
- L15:1280 weakly excites
- L27:2138 weakly suppresses
- L25:2623 weakly suppresses
- L5:418 weakly excites
- L9:1621 weakly suppresses
- L18:2238 weakly excites
- L14:659 weakly excites
- L12:2087 weakly suppresses
- L5:2799 weakly suppresses
- L9:1434 weakly suppresses
- L16:2448 weakly suppresses
- L25:2213 weakly suppresses
- L17:552 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
- 免
- mailto
- .setInput
- 就不
- ows
- feld
- ós
- 将其
- -Time
- act
Suppresses
- 道路上
- earer
- 瞎
- anthology
- ighest
- icked
- abyrin
- меща
- .localization
- UNDLE