Neuron Explorer

Qwen3-0.6B

full stops in running text

A neuron in layer 27 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 responds to full stops and dashes in ordinary prose, mostly where a sentence or clause ends. Its signals are much weaker than the other neurons here. Turning it down or up produced repeated filler words, so the steering does not show what it does.

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

    Trial to Unlock the Full Review⏎Why Lesson Planet?⏎Find quality lesson planning resources, fast!⏎Share & remix collections to collaborate.⏎Organize your curriculum with collections. Easy!⏎Have time to be more creative & energetic with your students!⏎Students are

  2. #2 peak

    Stop breadboarding and soldering – start making immediately! Adafruit’s Circuit Playground is jam-packed with LEDs, sensors, buttons, alligator clip pads and more. Build projects with Circuit Playground in a few minutes with the drag-and-drop MakeCode

  3. #3 peak

    have the proper application to view it first. In case of further problems read the IDEAS help page. Note that these files are not on the IDEAS site. Please be patient as the files may be large.⏎As the access to this document

  4. #4 peak

    zipping the .tgz files using the terminal. Let’s get into these.⏎Note: For this particular tutorial, we will use the Ubuntu 22.04. However, the commands used here are same for all the distributions.⏎

  5. #5 peak

    Trial to Unlock the Full Review⏎Why Lesson Planet?⏎Find quality lesson planning resources, fast!⏎Share & remix collections to collaborate.⏎Organize your curriculum with collections. Easy!⏎Have time to be more creative & energetic with your students!⏎Students explore

  6. #6 peak

    completely new experimental access to quantum physics has been made at University of Konstanz. The team of scientists headed by...⏎Yersiniae cause severe intestinal infections. Studies using Yersinia pseudotuberculosis as a model organism aim to elucid

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: in the in the in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in

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

    דמה DooMaH (Ezekiel27:32)⏎and דממה DiMaMaH (IKings19:12)⏎all mean silence. דמום DahMOOM is silent, דמם

  2. #2 peak

    under CC BY-NC-SA⏎Given an integer , the exponential sum version of the Gauss sum modulo is defined by⏎and the quadratic residue version is defined by⏎where is the Jacobi symbol, which has values

  3. #3 peak

    this problem?⏎offers the opportunity for a statistical investigation⏎of confidence intervals and expectations at various levels of⏎There are three possible levels at which this problem can be⏎tackled, and how it is used depends on the levels of the students

  4. #4 peak

    rendered that Aaron was DUMBFOUNDED. DIM⏎perception is seen in words like דמדום DiMDOOM⏎("dim light" and "blackout"-- as in unconscious). טמום

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

  1. L19:68 strongly excites full stops closing a claim
  2. L23:509 strongly suppresses
  3. L13:44 strongly suppresses inside web addresses and citations
  4. L22:3045 strongly suppresses
  5. L18:178 excites
  6. L25:30 suppresses
  7. L15:184 excites
  8. L26:27 excites
  9. L22:14 excites
  10. L15:390 excites
  11. L20:59 suppresses
  12. L11:223 excites
  13. L22:67 suppresses
  14. L23:10 excites
  15. L19:1240 suppresses
  16. L10:669 suppresses
  17. L19:214 excites
  18. L21:201 excites
  19. L15:59 suppresses
  20. L19:1174 excites
  21. L24:353 excites decimal point inside a number
  22. L19:124 suppresses
  23. L14:140 suppresses
  24. L10:155 excites

Feeds later layers

None: this is the last layer.

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.

1 values between -21.66 and -19.812 values between -12.43 and -10.582 values between -10.58 and -8.732 values between -8.73 and -6.893 values between -6.89 and -5.0418 values between -5.04 and -3.20767 values between -3.20 and -1.352652 values between -1.35 and 0.5095 values between 0.50 and 2.3469 values between 2.34 and 4.1949 values between 4.19 and 6.0436 values between 6.04 and 7.8828 values between 7.88 and 9.7323 values between 9.73 and 11.5725 values between 11.57 and 13.4210 values between 13.42 and 15.2711 values between 15.27 and 17.119 values between 17.11 and 18.966 values between 18.96 and 20.803 values between 20.80 and 22.654 values between 22.65 and 24.504 values between 24.50 and 26.341 values between 26.34 and 28.191 values between 28.19 and 30.038 values between 30.03 and 31.885 values between 31.88 and 33.7310 values between 33.73 and 35.5710 values between 35.57 and 37.426 values between 37.42 and 39.273 values between 39.27 and 41.1115 values between 41.11 and 42.968 values between 42.96 and 44.805 values between 44.80 and 46.658 values between 46.65 and 48.507 values between 48.50 and 50.342 values between 50.34 and 52.19offnegativepositiveoften

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. egrity
  2. stakes
  3. missions
  4. 不知不
  5. assadors
  6. plies
  7. 鲑
  8. hone
  9. reece
  10. DOCTYPE

Suppresses

  1. in
  2. for
  3. then
  4. that
  5. to
  6. according
  7. if
  8. otherwise
  9. the
  10. during