Neuron Explorer

Qwen3-0.6B

unclear: scattered mid-sentence tokens

A neuron in layer 7 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 ordinary words in the middle of sentences, such as 'at', 'the' and 'is', with no clear shared meaning. Turning it up made the model get stuck repeating a single word like 'She' or 'The'.

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

    station, municipal yard,⏎highway department, or school/library. There are some sites that⏎are operated by media such as newspapers, and a few that are at⏎ahead can help you locate just exactly where the station might⏎be

  2. #2 peak

    ⏎Minerals can be identified by their hardness,⏎colour, the way they reflect light (lustre), the⏎way they break (cleavage and fracture), and⏎Mineral crystals fall into six crystal systems,⏎each of which is

  3. #3 peak

    that it⏎bridges them. Using a straw, blow air underneath⏎the paper. This causes a downward force on the paper, and it⏎moves down in the center.⏎By blowing the air between

  4. #4 peak

    Kansas, Texas, New Mexico, Montana, Alaska, and California; Canada=⏎Manitoba, Alberta, British Columbia; Mexico; Brazil; Poland; Denmark=Bornholm,⏎Greenland; Austria; Switzerland; Hungary;

  5. #5 peak

    shapes in this example, labelled 1, 2 and 3 as follows:⏎2222.2222 ..22.2... 11.222..3 ....2..33⏎

  6. #6 peak

    .⏎The surface is level or slightly undulating near the lake, but in the E.⏎part rough and mountainous, and more adapted to grazing. The soil is fertile,⏎particularly along the lake shore. Indian corn, wool, oats, and

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 fact that, for example, the function f(x) = x^2 - 4x^2 + 6x^2 is the same as f(x) = 6x^2 - 4x^2.

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

    historical State of the Union Address to more contemporary ones. In this Government instructional activity, students compare and contrast historical Presidential speeches to the ones they hear today. High...⏎9th - 12th Social Studies & History⏎Anticipating the

  2. #2 peak

    ons⏎together? How will you know you have found all the ways?⏎Swap the stars with the moons, using only knights' moves (as on a⏎chess board). What is the smallest number of moves possible?⏎Can you work

  3. #3 peak

    this⏎Dig into the entry’s quotation paragraphs, and you’ll see that the OED

  4. #4 peak

    02, 2012⏎Repairing bones with nanomaterials⏎(Nanowerk News) Around

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. L4:156 strongly excites
  2. L4:224 excites
  3. L6:1374 excites
  4. L4:32 suppresses
  5. L5:1778 suppresses
  6. L1:2545 weakly excites
  7. L5:187 weakly suppresses
  8. L4:351 weakly suppresses
  9. L2:666 weakly suppresses
  10. L6:2586 weakly suppresses
  11. L4:138 weakly excites
  12. L6:191 weakly suppresses
  13. L5:238 weakly excites
  14. L1:1016 weakly suppresses
  15. L3:468 weakly suppresses
  16. L5:1240 weakly excites
  17. L2:673 weakly excites
  18. L6:203 weakly excites
  19. L6:2863 weakly suppresses
  20. L0:432 weakly suppresses
  21. L0:1403 weakly suppresses
  22. L1:1520 weakly excites
  23. L4:30 weakly excites
  24. L3:721 weakly excites

Feeds later layers

  1. L16:442 strongly suppresses
  2. L20:268 strongly excites
  3. L22:377 strongly excites
  4. L14:568 strongly excites
  5. L9:234 strongly suppresses line break inside hard-wrapped text
  6. L15:643 strongly excites
  7. L16:317 strongly excites
  8. L18:795 strongly excites
  9. L8:380 strongly suppresses
  10. L24:1619 strongly excites
  11. L26:1309 strongly excites
  12. L19:558 strongly excites
  13. L13:309 strongly excites
  14. L21:1164 strongly excites
  15. L25:2204 strongly excites
  16. L25:533 strongly excites first letter m or oct fragments
  17. L19:771 strongly excites
  18. L21:579 strongly excites
  19. L19:901 strongly excites
  20. L10:230 strongly excites
  21. L23:607 excites
  22. L16:795 excites
  23. L26:758 excites
  24. L27:947 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.

1 values between -5.45 and -5.281 values between -4.78 and -4.621 values between -4.28 and -4.121 values between -3.78 and -3.621 values between -3.12 and -2.952 values between -2.95 and -2.782 values between -2.78 and -2.621 values between -2.45 and -2.283 values between -2.28 and -2.122 values between -2.12 and -1.951 values between -1.95 and -1.781 values between -1.78 and -1.622 values between -1.62 and -1.454 values between -1.45 and -1.288 values between -1.28 and -1.1211 values between -1.12 and -0.9515 values between -0.95 and -0.7821 values between -0.78 and -0.6230 values between -0.62 and -0.4547 values between -0.45 and -0.28194 values between -0.28 and -0.122580 values between -0.12 and 0.05761 values between 0.05 and 0.22151 values between 0.22 and 0.3837 values between 0.38 and 0.5518 values between 0.55 and 0.726 values between 0.72 and 0.882 values between 0.88 and 1.054 values between 1.05 and 1.22offnegativepositiveoften

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. -⏎
  2. &⏎
  3. ,⏎
  4. ⏎
  5. ,⏎
  6. (⏎
  7. -⏎
  8. /⏎
  9. \⏎
  10. +⏎

Suppresses

  1. 县委书记
  2. ','',
  3. yat
  4. .'.$
  5. pewnością
  6. (*((
  7. ทำความสะอาด
  8. inely
  9. =").
  10. 联系我们