Neuron Explorer

Qwen3-0.6B

dots in initials and abbreviations

A neuron in layer 5 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

The neuron fires on the full stop that follows a single letter, like the dot in a person's middle initial ('A. M. Sullivan') or in 'M. D. Anderson'. Turning it up did not make the writing about names or initials in any obvious way.

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

  1. #1 peak

    0000000 SM: 11111111.11111110.00000000.00000000 IP: 1

  2. #2 peak

    Logic of Meaning and Meaningful Implication. B. Matalon(translated by J. Byrnes), A Genetic Study of Implication. R.J. Falmagne, Language and the Acquisition of Logical Knowledge. M.D.S.

  3. #3 peak

    on materials provided by NASA's Marshall Space Flight Center. Note: Materials may be edited for content and length.⏎Cite This Page:<|endoftext|>By A. M. Sullivan⏎CHAPTER IX. (continued)⏎From the Atlas and Cyclopedia of Ireland

  4. #4 peak

    d. Part V. Book IV. The Tissues and MechanNew edition in preparation. isms of Reproduction. Appendix. By A. S. LEA. 75. 6d. HANDBOOK OF

  5. #5 peak

    initial post(s) early in the module.⏎See the Discussion Board Guidelines and Rubrics⏎Download Discussion Board Guidelines and Rubrics.⏎Reh, F. J. (2018, November 01). Here Is a Look at

  6. #6 peak

    . Düchting). Computer simulation of the morphology and development of several species of seaweed using Lindenmayer systems (J.D. Corbit, D.J. Garbary). Generating fractals from Voronoi diagrams (K.W.

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: the number of, the number of, the number of, the number of, the number of, the number of, the number of, the number of, the number of, the number of, the number of, the number of,

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

    comparing & contrasting modalities (a piece of lit to a video), or may do well in geometry…. Graphic organizers are a great resource for them⏎👩👩👦👦 Some enjoy working in

  2. #2 peak

    the full content.⏎This preview has intentionally blurred sections. Sign up to view the full version.View Full Document⏎Unformatted text preview: 60 | | | | CHAPTER 1 FUNCTIONS AND MODELS Not all functions possess inverses. Lets

  3. #3 peak

    even need to be real words, take, for example, Hanson's "MMMBop" or The Police's "De Do Do Do, De Da Da Da." Use whatever works.⏎Make compromises. You can take pre-existing lyrics and fit

  4. #4 peak

    .0/15 IP: 10101100.00010000.00000000.00000000 SM: 111

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. L4:503 strongly suppresses
  2. L3:2985 strongly suppresses
  3. L4:330 strongly excites
  4. L3:1046 strongly excites
  5. L3:305 excites
  6. L1:2 suppresses
  7. L1:842 excites
  8. L0:7 excites
  9. L3:11 excites
  10. L2:2068 excites
  11. L2:878 suppresses
  12. L3:440 suppresses
  13. L0:14 excites
  14. L0:349 suppresses
  15. L0:508 suppresses
  16. L1:2115 suppresses
  17. L0:429 suppresses
  18. L0:617 excites
  19. L1:452 suppresses
  20. L4:235 weakly suppresses
  21. L2:47 weakly suppresses
  22. L0:732 weakly excites
  23. L0:188 weakly excites
  24. L0:1131 weakly excites

Feeds later layers

  1. L14:642 strongly suppresses
  2. L8:2572 strongly excites
  3. L9:1142 strongly suppresses
  4. L10:1091 strongly excites
  5. L13:1596 strongly excites
  6. L11:2666 strongly suppresses
  7. L11:1872 strongly suppresses
  8. L6:513 excites
  9. L11:1153 suppresses
  10. L6:1405 suppresses
  11. L17:1622 excites
  12. L15:344 suppresses
  13. L6:1839 suppresses
  14. L8:1435 excites
  15. L14:374 suppresses
  16. L9:146 suppresses
  17. L19:2663 suppresses
  18. L8:763 suppresses
  19. L12:1089 excites
  20. L10:85 suppresses
  21. L17:1590 suppresses
  22. L10:516 excites
  23. L13:1512 suppresses
  24. L12:1377 suppresses

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 -1.26 and -1.131 values between -0.73 and -0.601 values between -0.60 and -0.476 values between -0.47 and -0.3482 values between -0.34 and -0.21601 values between -0.21 and -0.081974 values between -0.08 and 0.061028 values between 0.06 and 0.19184 values between 0.19 and 0.3220 values between 0.32 and 0.453 values between 0.45 and 0.581 values between 0.58 and 0.711 values between 0.71 and 0.842 values between 0.84 and 0.981 values between 0.98 and 1.111 values between 1.37 and 1.501 values between 3.87 and 4.00offnegativepositiveoften

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. pedia
  2. เหน
  3. ewood
  4. 之上
  5. 而来
  6. 之事
  7. 而成
  8. xa
  9. 但是如果
  10. 怙

Suppresses

  1. ادات
  2. /cms
  3. doen
  4. /do
  5. renc
  6. UDO
  7. angu
  8. weeds
  9. 朋友们
  10. udio