Neuron Explorer

Qwen3-0.6B

full stop ending a heading or label

A neuron in layer 1 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 a full stop that closes a short heading or label, like "Topics." or "Eccl. Law." It goes quiet or negative on full stops and quote marks that end a full sentence. Turning it up or down did not clearly change 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

    . If a larger rope was needed, three ropes could be twisted together to form anchor cable.⏎Topics. This historical marker is listed in this topic list: Industry & Commerce.⏎Location.

  2. #2 peak

    brief, but when the facts are material they cannot be too numerous when the argument is pertinent and weighty, it cannot be too extended.⏎Eccl. Law. The name of a kind of papal rescript.<|endoftext|>Edward Hopper (

  3. #3 peak

    I. Book I. Blood: The Tissues of Movement, the Vascular Mechanism. Sixth Edition. 10s. 6d. Part II. Book II. The

  4. #4 peak

    be considered widespread, the pain must occur on both sides of your body and above and below your waist.⏎- Headache. Frequent headaches occur in a majority of children who have fibromyalgia.⏎

  5. #5 peak

    during this period of growth, such as annual warm-season grasses or legumes where possible. You might also consider feeding hay or grain.⏎De-stock pastures. Livestock pressure on pastures can be reduced by selling unproductive livestock,

  6. #6 peak

    own. This is a fabulous little book for parents and teachers to introduce children to the enchantment and magicality found in words. Highly recommended. 5 Stars. First reviewed for Readers Favorite.⏎by Fiona Ingram<|endoftext|>Fishing is both an art

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 how to be a good thing, and I have been thinking about how to be a good thing. Now, I have been thinking about how to be a good thing. Now, I have been thinking about how to

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

    findings were recently published in PNAS⏎Our ideas of protein function have undergone a marked transformation with the discovery of an aspect of enzyme evolution called "moonlighting." That is, a protein can acquire a second (moonlighting) function without

  2. #2 peak

    immediately set to work expanding that program and also establishing his family in a comfortable home on Cornell’s campus, along a stretch of East Avenue known as “Faculty Row.” The Bailey property became known as “Garden Home,” and the prefaces of

  3. #3 peak

    recent introduction of new APIs and tools for the Java Server Pages (JSP) environment, marrying XML data to Java-based services and applications is easier than ever before. The September 17 release of the JavaServer Pages 1.2 specification

  4. #4 peak

    goddess in Scandinavian was "Ostra," for Anglo-Saxons it was "Eostre," and the current region of Germany she was called "Eastre." See the

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. L0:1224 weakly suppresses
  2. L0:348 weakly excites space token right after a period
  3. L0:178 weakly suppresses
  4. L0:470 weakly suppresses
  5. L0:509 weakly suppresses
  6. L0:1301 weakly suppresses
  7. L0:308 weakly suppresses
  8. L0:2855 weakly suppresses
  9. L0:1369 weakly suppresses
  10. L0:1453 weakly suppresses
  11. L0:1567 weakly excites
  12. L0:71 weakly excites
  13. L0:627 weakly excites
  14. L0:1241 weakly suppresses
  15. L0:625 weakly suppresses
  16. L0:2408 weakly excites
  17. L0:2475 weakly suppresses
  18. L0:999 weakly excites
  19. L0:1576 weakly suppresses
  20. L0:299 weakly excites
  21. L0:951 weakly suppresses
  22. L0:1473 weakly suppresses
  23. L0:378 weakly suppresses
  24. L0:609 weakly suppresses 'of' in 'mode of'

Feeds later layers

  1. L4:80 suppresses
  2. L11:879 weakly suppresses
  3. L5:402 weakly suppresses
  4. L16:659 weakly suppresses
  5. L21:190 weakly suppresses
  6. L18:1085 weakly suppresses full stops ending sentences
  7. L15:764 weakly suppresses
  8. L18:442 weakly suppresses
  9. L19:619 weakly suppresses
  10. L23:614 weakly excites
  11. L10:295 weakly suppresses
  12. L19:185 weakly suppresses
  13. L5:46 weakly excites
  14. L25:66 weakly suppresses
  15. L18:90 weakly suppresses
  16. L11:955 weakly suppresses
  17. L7:972 weakly excites
  18. L3:621 weakly excites
  19. L3:141 weakly excites
  20. L5:186 weakly suppresses
  21. L24:73 weakly suppresses
  22. L22:1911 weakly excites
  23. L2:12 weakly excites
  24. L13:73 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.

2 values between -3.03 and -2.891 values between -2.89 and -2.761 values between -2.76 and -2.621 values between -2.49 and -2.353 values between -2.35 and -2.224 values between -2.22 and -2.084 values between -2.08 and -1.955 values between -1.95 and -1.813 values between -1.81 and -1.685 values between -1.68 and -1.546 values between -1.54 and -1.414 values between -1.41 and -1.274 values between -1.27 and -1.149 values between -1.14 and -1.008 values between -1.00 and -0.8711 values between -0.87 and -0.736 values between -0.73 and -0.606 values between -0.60 and -0.4616 values between -0.46 and -0.3388 values between -0.33 and -0.19650 values between -0.19 and -0.061779 values between -0.06 and 0.081063 values between 0.08 and 0.21205 values between 0.21 and 0.3514 values between 0.35 and 0.483 values between 0.48 and 0.621 values between 0.62 and 0.752 values between 0.89 and 1.022 values between 1.43 and 1.561 values between 1.83 and 1.971 values between 2.24 and 2.37offnegativepositiveoften

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. ippet
  4. nea
  5. orrh
  6. malink
  7. atég
  8. porto
  9. arness
  10. openh

Suppresses

  1. ---⏎⏎
  2. jadi
  3. зи
  4. =log
  5. ays
  6. כס
  7. 至此
  8. aying
  9. 围棋
  10. ["$