Neuron Explorer

Qwen3-0.6B

'and' inside lists

A neuron in layer 14 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 fires on the word 'and' right before the last item of a list, like 'books, notebooks, and' or 'panting and'. It is the list-joining 'and' that it cares about. Turning it down made the writing repeat 'and the' over and over.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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

    for TMJ.⏎How Are Temporomandibular Joint Disorders Treated?⏎There are a variety of treatment options for TMJ disorder, including physical therapy and surgery. Physical therapy involves relaxing overworked muscles and addressing any inflammation. Surgery may

  2. #2 peak

    system, and multithreading.⏎Also included is coverage of some of Java's most powerful features,⏎such as generics, lambda expressions, and the very important⏎modules feature. A number of recent additions to the Java language

  3. #3 peak

    may include testing of your peripheral vision (visual field test). You may have a partial or complete neurologic exam to test your strength, sensation, and coordination. The neuro-ophthalmologist will review the records and scans from previous

  4. #4 peak

    feed intake. When this is insufficient to keep their body temperature below 39.3°C, they exhibit symptoms of clinical heat stress, such as panting and drooling, and their welfare, health, and milk production can be severely affected

  5. #5 peak

    or a teenager, a lot of us have a backpack to use when going to school. These backpacks are used to store things such as books, notebooks, and

  6. #6 peak

    colleagues (Meltzer, Gatward, Corbin, Goodman, & Ford, 2003; Meltzer, Gatward, Goodman, & Ford, 2000). Our teacher ratings reflected similar levels of disturbance

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 have been thinking about the time, the time, the time, the time, the time, the time, the time, the time, the time, the time, and, and, and, and, and, and, and

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

    |Journal||Classical and Quantum Gravity|⏎|State||Published - Jul 7 2003|⏎ASJC Scopus subject areas⏎- Physics and Astronomy (miscellaneous)<|endoftext|>Most computer users are familiar with the concept of spy

  2. #2 peak

    in a discontinuity analysis on observations near the borders between anglophone and francophone countries.⏎- Colonial rule⏎ASJC Scopus subject areas⏎- Economics and Econometrics⏎- Organizational Behavior and Human Resource Management<|endoftext|>- About ASTI

  3. #3 peak

    text, Inventing the medium, J. Murray describes three layers of media design. These include inscription, transmission and representation. Inscription is the engraving of a message or a code on a

  4. #4 peak

    article. With respect to symmetries and invariance principles, three special conservation laws have been described, associated with inversion or reversal of space, time, and charge.⏎Conservation laws are considered to be fundamental laws of nature, with broad

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. L13:557 strongly suppresses
  2. L10:74 strongly excites
  3. L13:1296 strongly suppresses
  4. L13:1066 strongly excites
  5. L8:1605 suppresses
  6. L11:876 suppresses 'other' after 'among' in lists
  7. L13:131 excites 'and' joining parallel phrases
  8. L13:1734 excites sentence end right before a line break
  9. L9:92 suppresses "and" or "or" joining two items
  10. L10:667 excites commas inside lists of names
  11. L6:620 excites "and" in lists of three or more
  12. L5:548 excites
  13. L10:782 excites
  14. L8:1238 excites
  15. L7:270 excites
  16. L8:1047 excites
  17. L12:1440 suppresses
  18. L10:1643 suppresses
  19. L3:1718 suppresses
  20. L7:578 excites 'and' in lists of three
  21. L12:2811 suppresses
  22. L12:690 suppresses
  23. L8:986 excites
  24. L4:57 suppresses

Feeds later layers

  1. L16:2442 strongly excites
  2. L17:919 suppresses
  3. L15:2277 excites
  4. L26:347 suppresses the 're' in self-reliant and similar
  5. L16:1669 excites
  6. L20:1856 suppresses
  7. L18:486 excites
  8. L25:6 excites
  9. L16:2577 suppresses
  10. L15:132 excites
  11. L19:1369 suppresses
  12. L27:289 suppresses
  13. L17:2221 suppresses
  14. L24:205 suppresses
  15. L19:1605 suppresses
  16. L18:1207 suppresses
  17. L16:2432 suppresses
  18. L15:2765 excites
  19. L17:632 suppresses unclear: coordinating words in lists
  20. L20:882 suppresses
  21. L18:2691 suppresses
  22. L18:120 excites
  23. L26:65 excites
  24. L16:1119 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 -1.78 and -1.592 values between -1.59 and -1.401 values between -1.40 and -1.203 values between -1.20 and -1.014 values between -1.01 and -0.8112 values between -0.81 and -0.6232 values between -0.62 and -0.42270 values between -0.42 and -0.231665 values between -0.23 and -0.031634 values between -0.03 and 0.16110 values between 0.16 and 0.3632 values between 0.36 and 0.5519 values between 0.55 and 0.7519 values between 0.75 and 0.9415 values between 0.94 and 1.1412 values between 1.14 and 1.3310 values between 1.33 and 1.5311 values between 1.53 and 1.728 values between 1.72 and 1.9210 values between 1.92 and 2.113 values between 2.11 and 2.313 values between 2.31 and 2.505 values between 2.50 and 2.693 values between 2.69 and 2.893 values between 2.89 and 3.083 values between 3.08 and 3.283 values between 3.28 and 3.471 values between 3.47 and 3.671 values between 3.67 and 3.862 values between 3.86 and 4.061 values between 4.06 and 4.251 values between 4.25 and 4.451 values between 4.45 and 4.644 values between 4.64 and 4.842 values between 5.03 and 5.231 values between 5.81 and 6.00offnegativepositiveoften

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. embro
  2. ú
  3. owards
  4. омер
  5. 后者
  6. ITTLE
  7. inction
  8. ledged
  9. łę
  10. accordingly

Suppresses

  1. 和
  2. 和技术
  3. 和个人
  4. 和其他
  5. 和社会
  6. and
  7. 和他的
  8. và
  9. и
  10. 企业和