Neuron Explorer

Qwen3-0.6B

'long' in 'long-term' and 'long-lasting'

A neuron in layer 24 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 'long' when it starts a phrase like 'long-term' or 'long-lasting'. Turning it down made the model repeat 'long' over and over, so it seems tied to that word.

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

    of her last child may affect telomere length, ultimately impacting long-term health, the researchers said.⏎Longer telomeres are thought to be beneficial for long-term health, while shorter ones can signify “various chronic conditions such as cardiovascular

  2. #2 peak

    opportunity that would help sustain the museum long-term.⏎<|endoftext|>“A 200:1 ratio is very difficult to overcome

  3. #3 peak

    devices even come equipped with a back-up battery to supply instant hit once the battery dies. By not delivering nicotine, they’re not exposing one to medical risks of long-term

  4. #4 peak

    forestry and organic farming, Burung Indonesia (BirdLife in Indonesia) has worked with 27 villages around Mbeliling to improve the economic productivity and long-term viability of the local landscape. The soil is managed to avoid

  5. #5 peak

    with oxygen at the fuel cell's anode. Just as described in<|endoftext|>A new assessment finds that a majority of clean energy technologies are failing to keep pace with long-term climate,

  6. #6 peak

    be one of the largest in the area. For surrounding landowners, the project will promote different methods of long-term forest protection, such as conservation agreements, easements, and tax incentives such

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: long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long long

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

    or false according to the video. This general English lesson plan finishes with a short discussion on smiling and the speaker’s ideas.⏎Because the video is about 8 minutes long, you can watch it with your students during the lesson and do the comprehension

  2. #2 peak

    a simple protein⏎. It is composed of 3 long chains wound together in a tight triple helix⏎. Each chain is 1400 amino acids⏎long. A repeated sequence of the amino acids glycine⏎make up each

  3. #3 peak

    may be secured. The home vegetable garden is worthy of<|endoftext|>MicroRNAs (miRNAs) are single-stranded, 18–24 nucleotide long, non-coding RNA molecules. They are involved in virtually every cellular process

  4. #4 peak

    under age 18 from anywhere in the world. Each contestant can submit one name, up to 16 characters long. Entries must include a short explanation and rationale for the name. Submissions

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. L23:1490 strongly excites
  2. L23:2296 strongly suppresses
  3. L19:1712 excites
  4. L22:2704 excites
  5. L23:2238 suppresses
  6. L22:1122 suppresses
  7. L18:1641 excites the word 'long' in phrases
  8. L23:631 suppresses
  9. L10:852 excites
  10. L21:1628 excites
  11. L18:1847 excites
  12. L4:1978 excites
  13. L22:1820 suppresses
  14. L22:557 suppresses
  15. L10:1502 excites
  16. L18:2800 excites
  17. L22:2790 suppresses
  18. L7:1101 suppresses
  19. L23:2601 suppresses unclear: numbers and distance words
  20. L8:918 excites
  21. L21:658 excites
  22. L21:1177 excites
  23. L12:2377 excites
  24. L7:1114 suppresses

Feeds later layers

  1. L27:1178 strongly suppresses
  2. L26:1469 strongly suppresses
  3. L27:1469 suppresses
  4. L27:2280 suppresses
  5. L25:2963 weakly excites
  6. L25:2842 weakly suppresses the word a before long
  7. L26:1343 weakly excites
  8. L27:3038 weakly suppresses
  9. L26:2153 weakly excites
  10. L25:1456 weakly excites
  11. L25:432 weakly suppresses
  12. L26:2855 weakly suppresses
  13. L25:3037 weakly suppresses
  14. L26:1094 weakly suppresses
  15. L25:2424 weakly excites
  16. L27:1351 weakly excites
  17. L25:1808 weakly suppresses
  18. L25:2311 weakly excites
  19. L25:2823 weakly excites
  20. L27:1179 weakly excites
  21. L26:2610 weakly suppresses
  22. L25:1625 weakly excites
  23. L26:2025 weakly excites
  24. L26:2112 weakly 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 -8.86 and -8.101 values between -8.10 and -7.342 values between -7.34 and -6.593 values between -6.59 and -5.837 values between -5.83 and -5.077 values between -5.07 and -4.3113 values between -4.31 and -3.5514 values between -3.55 and -2.7930 values between -2.79 and -2.0466 values between -2.04 and -1.28329 values between -1.28 and -0.522550 values between -0.52 and 0.24590 values between 0.24 and 1.00130 values between 1.00 and 1.7657 values between 1.76 and 2.5137 values between 2.51 and 3.2721 values between 3.27 and 4.0313 values between 4.03 and 4.798 values between 4.79 and 5.5513 values between 5.55 and 6.312 values between 6.31 and 7.064 values between 7.06 and 7.821 values between 7.82 and 8.581 values between 8.58 and 9.341 values between 9.34 and 10.101 values between 10.10 and 10.851 values between 10.85 and 11.611 values between 11.61 and 12.371 values between 13.13 and 13.892 values between 19.19 and 19.951 values between 20.71 and 21.47offnegativepositiveoften

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. ivial
  2. ifice
  3. ücü
  4. entious
  5. ṇ
  6. ernaut
  7. astery
  8. nex
  9. oge
  10. heit

Suppresses

  1. long
  2. long
  3. 长
  4. Long
  5. Long
  6. →long
  7. LONG
  8. longer
  9. (long
  10. .long