Neuron Explorer

Qwen3-0.6B

unclear: mixed word fragments and phrases

A neuron in layer 27 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on a grab bag of word pieces and phrases with no clear shared theme. When it was pushed down, the model kept writing the word 'string'. When pushed up, it got stuck repeating words.

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

    be great for routine maintenance, but they may not have the knowledge to apply the correct gardening principals to your more complex gardening needs.⏎Common pruning mistake! The uniform shearing of shrubs may be more pleasing to the eye for some but, as

  2. #2 peak

    God and is profitable for doctrine, for reproof, for correction, for instruction in righteousness: That the man of God may

  3. #3 peak

    virus.⏎The hepatitis B vaccine consists of a small protein from the surface of the hepatitis B virus called the hepatitis B surface antigen

  4. #4 peak

    section 3: Extended Library Functions, Volume 2 Oracle Solaris 11.1 Information Library|⏎- load exponent of a floating point number⏎c99 [ flag... ] file... -lm [ library...

  5. #5 peak

    had been there; apparently he planned on it, but he is ill. He has spearheaded a remarkable project that has united the fields of homotopy theory, mathematical logic, and the theory of programming languages — and in the process, it

  6. #6 peak

    1 , stored in file hw3_ex1.asm , that prompts the user to type a 5-character string. The program reads the first 5 characters of that string and prints two strings. The first string contains the 5 characters entered by

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: string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string string

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

    Angle Sum⏎Bit to Gigabyte⏎Find the treasures in MATLAB Central and discover how the community can help you!⏎Choose a web site to get translated content where available and

  2. #2 peak

    the pagan message concealed in the thought of the contemporary Greek Platonist Georgios Gemistos Pletho and warned of the danger to orthodoxy if a Platonist like Bessarion were ever elected pope. The book precipitated

  3. #3 peak

    ak Dun v. Yukon, 2015 YKCA 18 (the “appeal decision”), the Yukon Court of Appeal unanimously allowed the Yukon government’s

  4. #4 peak

    only provides valuable insights into an organization and its customers, but it empowers teams to take action.⏎However, incomplete or inaccurate data can bring it all to a screeching halt.⏎The importance of clean data⏎Clean data — data that is

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. L24:905 strongly suppresses
  2. L26:229 excites
  3. L23:2751 suppresses
  4. L23:455 suppresses
  5. L24:479 excites
  6. L22:1136 weakly excites
  7. L13:1469 weakly suppresses
  8. L22:1920 weakly suppresses
  9. L15:1656 weakly suppresses
  10. L23:1647 weakly excites
  11. L1:1541 weakly excites
  12. L21:546 weakly suppresses
  13. L21:225 weakly excites
  14. L10:886 weakly excites
  15. L5:216 weakly suppresses
  16. L9:878 weakly excites
  17. L1:2027 weakly excites
  18. L14:2520 weakly suppresses
  19. L17:265 weakly suppresses
  20. L24:3065 weakly suppresses
  21. L15:1203 weakly excites
  22. L12:1131 weakly excites
  23. L10:32 weakly excites
  24. L18:1016 weakly suppresses

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 -18.34 and -17.563 values between -5.79 and -5.004 values between -5.00 and -4.224 values between -4.22 and -3.437 values between -3.43 and -2.6518 values between -2.65 and -1.8645 values between -1.86 and -1.08302 values between -1.08 and -0.292873 values between -0.29 and 0.49547 values between 0.49 and 1.2848 values between 1.28 and 2.0620 values between 2.06 and 2.858 values between 2.85 and 3.635 values between 3.63 and 4.416 values between 4.41 and 5.201 values between 5.20 and 5.983 values between 5.98 and 6.777 values between 6.77 and 7.553 values between 7.55 and 8.341 values between 10.69 and 11.481 values between 11.48 and 12.261 values between 12.26 and 13.05offnegativepositiveoften

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. erton
  2. ��
  3. atically
  4. мент
  5. feared
  6. 国安
  7. inson
  8. otten
  9. ации
  10. achts

Suppresses

  1. (string
  2. string
  3. (vector
  4. string
  5. →string
  6. vector
  7. ,string
  8. vector
  9. (object
  10. ,vector