Neuron Explorer

Qwen3-0.6B

'between' after link or relationship nouns

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 'between' when it follows words like 'link', 'connection', 'correspondence' or 'association'. Turning it up did not produce a clear change in what the model wrote.

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

    studies have delivered inconsistent results in attempting to determine whether early gluten consumption increases the likelihood of developing celiac disease. Previously conducted studies have found that there is no connection between early consumption of gluten, and even that late introduction of gluten may be a contributing

  2. #2 peak

    is Noether's theorem, which states that there is a one-to-one correspondence between each one of them and a differentiable symmetry of nature. For example, the

  3. #3 peak

    no one had answered in this way."⏎In their paper, published in the current issue of the Public Library of Science ONE, they say there's an established link between Type 1 diabetes and the body's ability to

  4. #4 peak

    of any kind, not just dark) were also less likely to report depressive symptoms than those who didn’t eat chocolate at all. However researchers found no significant link between any non‐dark chocolate consumption and clinically relevant depressive symptoms.⏎“This study provides some

  5. #5 peak

    experience upset stomach, headaches, and increased heart rate and blood pressure. The products that contain caffeine often cause other problems, too. Studies have found a significant connection between consumption of sugary drinks and weight gain in children. Stick to water and milk

  6. #6 peak

    <|endoftext|>Myth 1: There has been no reduction in poverty as a result of post-1991 reforms⏎This myth challenges not only the link between reforms and growth but the link between growth and poverty, by asserting that poverty did

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 thing that I have been thinking about, the thing that I have been thinking about, the thing that I have been thinking about, the thing that I have been thinking about, the thing that I have been thinking about, the thing that

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

    52°30'58"N⏎Longitude: -3.0604 / 3°3'37"W⏎OS Eastings: 328136⏎OS Northings: 291

  2. #2 peak

    东方在线 发布人:young⏎14. Energy Resources⏎China is rich in various energy resources.

  3. #3 peak

    00) ⇒ #<marker at 100 in markers.texi>⏎;; Now insert one character at the beginning of the buffer: (goto-char (point-min)) ⇒ 1 (insert "Q") ⇒ nil⏎<|endoftext|>A

  4. #4 peak

    /var/www/html/drupal⏎# cp sites/default/default.settings.php sites/default/settings.php⏎Now make a “default” directory writable by installer.⏎# chmod -R a+w sites/default/⏎Step 4: Create the MySQL Database⏎I

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. L12:1378 excites
  2. L10:2874 suppresses
  3. L12:2891 excites
  4. L13:3061 suppresses
  5. L10:712 suppresses
  6. L13:2514 suppresses
  7. L12:1589 suppresses
  8. L9:2827 suppresses
  9. L10:1836 weakly suppresses
  10. L12:277 weakly excites
  11. L13:2294 weakly excites
  12. L13:2723 weakly suppresses
  13. L12:1502 weakly excites
  14. L3:2871 weakly suppresses
  15. L11:1039 weakly suppresses
  16. L0:1739 weakly excites
  17. L5:1704 weakly excites
  18. L13:2521 weakly suppresses
  19. L6:402 weakly suppresses
  20. L1:2759 weakly excites
  21. L11:769 weakly suppresses
  22. L13:197 weakly excites
  23. L13:584 weakly excites unit conversions in parentheses
  24. L13:417 weakly excites

Feeds later layers

  1. L15:1172 excites
  2. L17:1779 excites
  3. L17:1488 excites
  4. L15:1159 suppresses ampersand in citation author lists
  5. L18:1401 excites
  6. L16:2378 excites
  7. L17:2040 excites
  8. L16:2329 excites
  9. L15:1679 weakly suppresses
  10. L15:3040 weakly suppresses
  11. L23:929 weakly suppresses
  12. L23:1020 weakly excites
  13. L19:2423 weakly suppresses
  14. L20:669 weakly suppresses
  15. L16:2950 weakly suppresses
  16. L18:1767 weakly excites
  17. L19:2683 weakly excites
  18. L16:1050 weakly excites
  19. L25:590 weakly suppresses
  20. L25:2825 weakly excites
  21. L24:2353 weakly excites
  22. L18:1955 weakly suppresses
  23. L26:2012 weakly excites
  24. L21:699 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 -0.90 and -0.831 values between -0.64 and -0.581 values between -0.58 and -0.515 values between -0.51 and -0.4517 values between -0.45 and -0.3918 values between -0.39 and -0.3266 values between -0.32 and -0.2697 values between -0.26 and -0.20256 values between -0.20 and -0.13399 values between -0.13 and -0.07576 values between -0.07 and -0.01700 values between -0.01 and 0.06619 values between 0.06 and 0.12460 values between 0.12 and 0.18340 values between 0.18 and 0.25192 values between 0.25 and 0.3185 values between 0.31 and 0.3849 values between 0.38 and 0.4416 values between 0.44 and 0.505 values between 0.50 and 0.572 values between 0.57 and 0.632 values between 0.88 and 0.951 values between 1.58 and 1.65offnegativepositiveoften

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. sembles
  2. Goodman
  3. 运气
  4. 罄
  5. Herrera
  6. setattr
  7. restrial
  8. cala
  9. vided
  10. ondo

Suppresses

  1. yl
  2. 夫
  3. иров
  4. ilst
  5. ér
  6. ��
  7. 是为了
  8. .are
  9. ume
  10. wend