Neuron Explorer

Qwen3-0.6B

line breaks after numbers or fragments

A neuron in layer 2 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 line break that comes after a number or a short, cut-off fragment at the end of a line, often in lists, page numbers or headings. Its negative side is weaker and falls on words like 'day' and 'boating'. Turning it up did not produce a clear theme in the written text.

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

    1,2⏎Whether it be driving the kids to extracurriculars, getting up at

  2. #2 peak

    University, USA⏎"TEEB brings a rigorous economic focus to bear on the problems of ecosystem degradation

  3. #3 peak

    ahead of formal qualifications’. 1⏎Research by the Institute of Leadership and Management with Ashridge Business School suggests that over

  4. #4 peak

    half, and half again, to⏎make four smaller rectangles. How many different ways can you fold⏎This problem focuses

  5. #5 peak

    On the other side, they⏎should write one sentence about Abraham Lincoln's life. After the penny is⏎created

  6. #6 peak

    3⏎Scientists learning what it takes for seals to live in the wilds of Antarctica⏎

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 told that I have a lot of information, and I have been told that I have a lot of information. What is the answer to this? - I have been told that I have a lot of information. What is the

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

    <|endoftext|>For most of his writing career, Vladimir Nabokov was frustrated by both the Freudian and behavorialist schools of psychology that reigned in his day, seeking a richer and more complex view of the mind than either provided. In

  2. #2 peak

    was replaced by the St. Lawrence Seaway which opened in 1959. In 2002, Lachine was reopened as a boating recreational area. An environmental reclamation project continues to clean up the polluted canal.⏎

  3. #3 peak

    would do<|endoftext|>Printed on a large, single sheet of paper (thus ‘broadside’), these cheap, everyday publications were the popular songs of their day, reaching the height of their popularity in the 1660s.

  4. #4 peak

    sheriff’s department, registry of deeds, and courthouses.⏎Portland harbor is a working waterfront serving tankers, tourists, and fisheries, along with recreational boating from year-round and summer residents. Ferries connect the mainland to the many Cas

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. L0:37 weakly excites
  2. L0:886 weakly excites
  3. L0:1224 weakly suppresses
  4. L0:592 weakly excites
  5. L1:1861 weakly excites
  6. L1:2791 weakly excites
  7. L0:1329 weakly excites
  8. L1:1277 weakly suppresses
  9. L0:178 weakly suppresses
  10. L0:2270 weakly suppresses
  11. L1:1190 weakly excites
  12. L0:268 weakly suppresses
  13. L1:1125 weakly suppresses
  14. L0:2432 weakly excites
  15. L1:2066 weakly suppresses
  16. L0:1243 weakly excites
  17. L0:267 weakly excites
  18. L1:1574 weakly excites
  19. L1:1537 weakly excites
  20. L0:532 weakly excites
  21. L1:1249 weakly suppresses
  22. L1:1511 weakly suppresses
  23. L0:1564 weakly excites
  24. L1:2986 weakly excites

Feeds later layers

  1. L19:902 excites
  2. L3:1353 weakly suppresses
  3. L4:795 weakly suppresses
  4. L18:1217 weakly suppresses
  5. L12:2055 weakly suppresses
  6. L27:387 weakly suppresses 'a' right after 'played' or 'play'
  7. L3:75 weakly suppresses
  8. L22:357 weakly excites
  9. L25:1644 weakly suppresses
  10. L18:2851 weakly suppresses
  11. L24:272 weakly suppresses
  12. L22:152 weakly excites
  13. L10:170 weakly excites
  14. L23:754 weakly suppresses
  15. L7:1960 weakly suppresses
  16. L4:547 weakly suppresses
  17. L27:2799 weakly suppresses the letter d inside words
  18. L21:2311 weakly excites
  19. L26:1063 weakly suppresses
  20. L27:1970 weakly suppresses
  21. L8:3064 weakly suppresses
  22. L7:2871 weakly suppresses
  23. L27:1749 weakly suppresses hyphen joining word parts
  24. L21:995 weakly 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 -1.21 and -1.151 values between -0.93 and -0.872 values between -0.87 and -0.823 values between -0.82 and -0.761 values between -0.65 and -0.594 values between -0.59 and -0.543 values between -0.54 and -0.488 values between -0.48 and -0.4211 values between -0.42 and -0.3736 values between -0.37 and -0.3143 values between -0.31 and -0.2666 values between -0.26 and -0.20112 values between -0.20 and -0.14236 values between -0.14 and -0.09497 values between -0.09 and -0.031565 values between -0.03 and 0.03739 values between 0.03 and 0.08338 values between 0.08 and 0.14117 values between 0.14 and 0.1957 values between 0.19 and 0.2524 values between 0.25 and 0.3110 values between 0.31 and 0.3610 values between 0.36 and 0.426 values between 0.42 and 0.471 values between 0.53 and 0.593 values between 0.59 and 0.641 values between 0.64 and 0.706 values between 0.70 and 0.762 values between 0.76 and 0.813 values between 0.81 and 0.871 values between 0.87 and 0.921 values between 0.98 and 1.04offnegativepositiveoften

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. imagine
  2. ide
  3. "`⏎⏎
  4. ment
  5. мен
  6. ife
  7. _Err
  8. Apache
  9. 供图
  10. owl

Suppresses

  1. ila
  2. quina
  3. enty
  4. esus
  5. →pr
  6. illard
  7. ByUsername
  8. venience
  9. าย
  10. ILA