Neuron Explorer

Qwen3-0.6B

periods ending informational sentences

A neuron in layer 18 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 lights up on the full stop at the end of a plain informational sentence, like the end of a sentence explaining a topic in an article. When it was turned up, the model's text drifted into rambling repetition, and when turned down it tended toward quiz-style answer choices.

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

    growth⏎Coffee is exclusively beneficial for hair health<|endoftext|>Ninjas vs Superbugs⏎Antibiotic-resistant bacteria - or, superbugs - are a growing threat. And MRSA is one of the worst. Learn more about Ninja Polymers

  2. #2 peak

    tgz File Using the Terminal?⏎If you are using open-source packages, then sooner or later, you will encounter the .tgz or .tar.gz files. This is a variation of zipped files; most open-source packages use this format

  3. #3 peak

    on the cross tabulation feature in Survey Analytics. Cross tabulations are a mainstay in the market research industry and is one of the most widely-used statistical tools. Cross

  4. #4 peak

    sports, in the NERF Camp⏎Probably one of the most fun activities where little ones can do a lot of movement playing outdoors is NERF camp. Here, children have the opportunity to play and bond with friends with other 3

  5. #5 peak

    most popular cryptocurrencies which can be used in 2020.⏎Maybe you’ve heard the terms digital gold or halving if you’ve been listening close enough. More importantly, you may have even heard about how Bitcoin or BTC prices have been

  6. #6 peak

    and stroke, among others using the same procedure.⏎MRI scans are painless and, unlike X-ray scans, they avoid X-ray radiation exposure, notes MedicineNet. There are

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 key point, as for the number, is to make the correct number. So, when I have been trying to think about the correct number, it is clear that the statement, "x" for the number, is the correct one

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

    human activities can theoretically exert some influence.(7) As the 21st century began, some experts thought it plausible

  2. #2 peak

    collection of wrought iron rivets, saddles and armature bars that make the statue structurally sound, and a double helix metal staircase is centered among them./Jared Owen<|endoftext|>New Study Issues Big Warning On Global Obesity Rates…⏎Published in

  3. #3 peak

    , freedom for the oppressed, and abundance for the poor. Therefore, the people waited for the Messiah or Christ.⏎The word Christmas, or "Christes Maesse," was first used in

  4. #4 peak

    labor. This absence of any higher comprehensive social philosophy led to efforts to supplement it by providing an overall philosophy of the public good. This absence led to Progressivism.At first it was a diverse movement of many who thought of themselves as “progressives

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. L17:944 suppresses
  2. L16:489 suppresses
  3. L15:2341 suppresses
  4. L17:2017 excites
  5. L17:1426 suppresses
  6. L2:156 excites
  7. L5:3 suppresses
  8. L9:12 suppresses
  9. L4:7 excites
  10. L14:1187 suppresses
  11. L9:2447 excites
  12. L16:2857 excites
  13. L17:990 excites
  14. L9:7 excites
  15. L15:1260 excites
  16. L16:1112 weakly suppresses
  17. L1:130 weakly suppresses
  18. L8:3048 weakly excites
  19. L10:33 weakly suppresses
  20. L12:731 weakly excites
  21. L15:2811 weakly excites
  22. L16:432 weakly suppresses period after month or initial
  23. L1:1 weakly excites
  24. L9:103 weakly excites "as" or "since" giving a reason

Feeds later layers

  1. L20:1147 excites comma after a sentence subject
  2. L20:1799 weakly suppresses
  3. L24:1720 weakly excites
  4. L20:2998 weakly suppresses
  5. L27:2653 weakly suppresses
  6. L19:1105 weakly suppresses
  7. L19:1485 weakly suppresses
  8. L19:2548 weakly suppresses
  9. L23:1875 weakly suppresses
  10. L26:1141 weakly excites
  11. L26:2170 weakly excites
  12. L19:297 weakly suppresses
  13. L21:555 weakly excites
  14. L21:3059 weakly suppresses 'the' after words like 'underscores'
  15. L24:3070 weakly suppresses
  16. L19:2370 weakly excites
  17. L27:1182 weakly excites
  18. L20:448 weakly excites
  19. L19:1656 weakly excites
  20. L25:2942 weakly excites
  21. L22:688 weakly suppresses
  22. L24:2307 weakly excites
  23. L20:1806 weakly excites slash inside web addresses
  24. L26:189 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 -12.31 and -11.721 values between -11.72 and -11.141 values between -10.55 and -9.973 values between -9.97 and -9.392 values between -9.39 and -8.802 values between -8.80 and -8.221 values between -8.22 and -7.641 values between -7.64 and -7.055 values between -7.05 and -6.474 values between -6.47 and -5.897 values between -5.89 and -5.304 values between -5.30 and -4.722 values between -4.72 and -4.149 values between -4.14 and -3.5514 values between -3.55 and -2.9712 values between -2.97 and -2.3931 values between -2.39 and -1.8044 values between -1.80 and -1.22113 values between -1.22 and -0.64783 values between -0.64 and -0.052482 values between -0.05 and 0.53197 values between 0.53 and 1.1159 values between 1.11 and 1.7042 values between 1.70 and 2.2817 values between 2.28 and 2.8619 values between 2.86 and 3.4514 values between 3.45 and 4.0310 values between 4.03 and 4.617 values between 4.61 and 5.202 values between 5.20 and 5.782 values between 5.78 and 6.366 values between 6.36 and 6.953 values between 6.95 and 7.533 values between 7.53 and 8.112 values between 8.11 and 8.702 values between 10.45 and 11.03offnegativepositiveoften

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. помощи
  3. |m
  4. acking
  5. }}/
  6. <Action
  7. 付费
  8. #undef
  9. ặng
  10. 掩

Suppresses

  1. ⏎⏎⏎
  2. ⏎ ⏎
  3. aroo
  4. .
  5. entre
  6. Grad
  7. ta
  8. Liqu
  9. scious
  10. sc