Neuron Explorer

Qwen3-0.6B

unclear: asterisk marker after a sentence

A neuron in layer 0 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron weakly reacts to a star or asterisk mark that comes right after a sentence or in code, such as a bullet or a pointer. The signal is faint and the pattern is not clear, and turning it up did not show a clear effect.

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 activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

  1. #1 peak

    the teacher and classroom through positive reinforcement, high expectations and discipline instills independence and enhances students growth. * IMPORTANCE OF CLASSROOM MANAGEMENT :- Effective classroom management paves the way for the

  2. #2 peak

    effectively removed from MB, Connecticut, and Rhode Island.⏎The Chesapeake: Bacon’s Rebellion and the Anglo-Powhatan Wars are the second supporting idea. * Nathaniel Bacon, an aristocrat in western Virginia and member of House of Burg

  3. #3 peak

    enough to be trees. This may take a year or two.⏎There's nothing "GMO" about this. These are breeding techniques and processes that were developed *before* the idea of genetic engineering which is the biochemical creation of specific advantageous mutations

  4. #4 peak

    Encomienda, the New England Confederation, and the Fur Trade, Dutch and French.⏎New England: King Phillip’s War is the first supporting idea. * The New England Confederation put to the test during war with Indian chieft

  5. #5 peak

    what I'm trying to do with those declarations. Humans read code, and humans infer things from the code that will be necessary to work on it.⏎typedef void * model_3d_ref;⏎model_3d_ref load_model_from_disk(const

  6. #6 peak

    (S.W.), the New England Confederation and the Fur Trade Dutch/French (N.Y. &N.F.) is the third supporting idea. * Amerindians forced to do unpaid labor to build roads, buildings, and other

Steering

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.

turned down a lot
Prompt

Here is something I have been thinking about lately: I am thinking of the problem of the type of the car, and the car's performance, and the car's safety and maintenance, and the car's design, and the car's price, and the car's environmental and safety standards,

Most negative on

Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.

  1. #1 peak

    00) ⇒ #<marker at 100 in markers.texi>⏎;; Now insert one character at the beginning of the buffer: (goto-char (point

  2. #2 peak

    as arguments. In such cases, the marker⏎stands for its current position.⏎Here are examples of creating markers, setting markers, and moving point to markers:⏎;; Make a new marker that initially does not point anywhere: (setq m1 (

  3. #3 peak

    licensed health care professional who diagnoses, treats, operates and prescribes for any disease,

  4. #4 peak

    . The regiment was mustered out in June 1865 at Nashville, and returned to Indianapolis.⏎Note Author: Shauna Borger<|endoftext|>|Product #: EMC5600046_TQ|⏎

Connections

Show on map

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Fed by earlier layers

None: this is the first layer.

Feeds later layers

  1. L4:701 weakly excites
  2. L18:2012 weakly excites
  3. L22:2035 weakly excites
  4. L16:1381 weakly excites
  5. L17:1833 weakly suppresses
  6. L20:2637 weakly suppresses
  7. L14:1821 weakly suppresses
  8. L21:349 weakly excites
  9. L5:1592 weakly excites
  10. L20:1571 weakly suppresses
  11. L3:630 weakly excites
  12. L26:1763 weakly excites
  13. L3:2407 weakly excites
  14. L8:2654 weakly excites
  15. L1:2197 weakly excites
  16. L1:1361 weakly excites
  17. L2:2119 weakly excites
  18. L4:791 weakly suppresses
  19. L15:639 weakly excites
  20. L5:2889 weakly excites
  21. L15:1976 weakly suppresses
  22. L10:2022 weakly excites
  23. L3:20 weakly suppresses
  24. L27:2403 weakly excites

Activation histogram

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.

2 values between -0.27 and -0.263 values between -0.24 and -0.222 values between -0.22 and -0.213 values between -0.20 and -0.189 values between -0.18 and -0.179 values between -0.17 and -0.1614 values between -0.16 and -0.1523 values between -0.15 and -0.1332 values between -0.13 and -0.1246 values between -0.12 and -0.1173 values between -0.11 and -0.0997 values between -0.09 and -0.08151 values between -0.08 and -0.07173 values between -0.07 and -0.06253 values between -0.06 and -0.04347 values between -0.04 and -0.03400 values between -0.03 and -0.02451 values between -0.02 and -0.00441 values between -0.00 and 0.01326 values between 0.01 and 0.02270 values between 0.02 and 0.03207 values between 0.03 and 0.05147 values between 0.05 and 0.06111 values between 0.06 and 0.0765 values between 0.07 and 0.0978 values between 0.09 and 0.1042 values between 0.10 and 0.1147 values between 0.11 and 0.1220 values between 0.12 and 0.1423 values between 0.14 and 0.1511 values between 0.15 and 0.1615 values between 0.16 and 0.186 values between 0.18 and 0.192 values between 0.19 and 0.204 values between 0.20 and 0.213 values between 0.21 and 0.232 values between 0.23 and 0.24offnegativepositiveoften

Logit lens

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 layers instead.

Promotes

  1. reiben
  2. abbo
  3. 半月
  4. 大会上
  5. oped
  6. ês
  7. caliente
  8. FormsModule
  9. ündig
  10. icum

Suppresses

  1. [
  2. odel
  3. components
  4. segment
  5. **
  6. second
  7. invariant
  8. -stack
  9. powers
  10. **⏎