Neuron Explorer

Qwen3-0.6B

question mark ending a web question heading

A neuron in layer 19 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 question mark at the end of a question-style heading, the kind often seen in FAQ pages and article titles, right before the answer begins. Turning it up gave off-topic or number-filled text, so the steering does not show this pattern.

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

    You might also be interested in:⏎What do smog, acid rain, carbon monoxide, fossil fuel exhausts, and tropospheric ozone have in common? They are all examples of air pollution.<|endoftext|>Revealing the art and architecture

  2. #2 peak

    culture.⏎This resource is licensed by LoubeeSav⏎under the iSLCollective Copyright License<|endoftext|>|Does a 'proper' tragic hero need a crown?|⏎Christmas Eve brought an email inviting me to write an essay about tragedy and social

  3. #3 peak

    selection of participants⏎- Type of participants⏎- democratic innovation yields a non-binding decision<|endoftext|>What does ‘diff’ mean in Git? A diff takes two data sets and shows you what has changed between them. Data

  4. #4 peak

    <|endoftext|>Which Joe gave his name to ‘sloppy joes’? We look at five interesting sandwiches and their lexical origins.⏎A small North American diving

  5. #5 peak

    now have customized climate predictions, thanks to a<|endoftext|>What is redistricting and how does it work in Iowa? Your questions answered.⏎The U.S. Census Bureau delivered its long-delayed population

  6. #6 peak

    or public awareness can be developed;⏎- Economic and Cultural Issues – Most of the efforts<|endoftext|>Doesn’t it seem like food allergies are way more common these days? When I was a kid, nobody worried about serving peanut butter at school, and

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:::::::::::::::::::::::::::::::::::::::::::::::::

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

    berg says.⏎Think fat just hangs around and does nothing? It doesn’t⏎By Nanci Hellmich, USA TODAY⏎PHOENIX — Most people think of fat as an inert blob, but fat cells release powerful

  2. #2 peak

    , more than 3,800 among adults 55 to 64 and around 800 among adults older than 65 — they accounted for a smaller proportion of deaths in these populations.⏎Among the eldest

  3. #3 peak

    , 2012⏎Share Email Print⏎Kate Wong is an editor and writer at Scientific American covering paleontology, archaeology and life sciences. Follow on Twitter⏎PALEO DIET: Analyses of tartar on the teeth

  4. #4 peak

    All intellectual property rights to the game are owned by respective owners in the U.S.A and Canada and the rest of the world. Anagrammer.com is not affiliated with Scrabble. This site is an educational tool

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. L18:2582 excites
  2. L18:5 excites
  3. L17:829 weakly excites
  4. L4:221 weakly suppresses
  5. L18:2043 weakly excites
  6. L14:392 weakly suppresses
  7. L2:204 weakly suppresses
  8. L6:718 weakly excites
  9. L16:32 weakly excites
  10. L2:7 weakly excites
  11. L18:887 weakly excites
  12. L16:118 weakly suppresses
  13. L9:7 weakly excites
  14. L17:38 weakly suppresses
  15. L14:51 weakly excites colon that introduces a statement
  16. L9:1109 weakly suppresses
  17. L18:2455 weakly excites
  18. L18:106 weakly excites
  19. L18:429 weakly excites
  20. L18:156 weakly excites
  21. L6:2486 weakly excites
  22. L14:1067 weakly excites
  23. L6:2678 weakly suppresses
  24. L9:366 weakly excites

Feeds later layers

  1. L21:189 suppresses document break token mid-sentence
  2. L20:1313 suppresses
  3. L22:541 excites
  4. L26:130 weakly excites ellipsis cut-off in quotes
  5. L25:273 weakly excites dollar sign before an amount
  6. L27:1053 weakly suppresses
  7. L22:416 weakly excites commas in dates and number lists
  8. L20:353 weakly excites
  9. L27:111 weakly suppresses
  10. L20:79 weakly suppresses
  11. L24:94 weakly suppresses
  12. L23:35 weakly excites
  13. L20:777 weakly suppresses
  14. L27:979 weakly suppresses
  15. L23:1878 weakly suppresses
  16. L20:398 weakly excites
  17. L25:360 weakly excites
  18. L26:56 weakly excites
  19. L24:2482 weakly suppresses
  20. L20:1710 weakly suppresses
  21. L21:1086 weakly excites end of sentence before a claim or statistic
  22. L21:50 weakly excites comma after 'So' or intro phrase
  23. L20:59 weakly suppresses
  24. L20:346 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.

3 values between -5.59 and -4.761 values between -4.76 and -3.947 values between -3.94 and -3.116 values between -3.11 and -2.2814 values between -2.28 and -1.4547 values between -1.45 and -0.622591 values between -0.62 and 0.211062 values between 0.21 and 1.0436 values between 1.04 and 1.8727 values between 1.87 and 2.708 values between 2.70 and 3.538 values between 3.53 and 4.365 values between 4.36 and 5.196 values between 5.19 and 6.028 values between 6.02 and 6.858 values between 6.85 and 7.687 values between 7.68 and 8.5010 values between 8.50 and 9.338 values between 9.33 and 10.163 values between 10.16 and 10.999 values between 10.99 and 11.824 values between 11.82 and 12.656 values between 12.65 and 13.481 values between 13.48 and 14.317 values between 14.31 and 15.145 values between 15.14 and 15.973 values between 15.97 and 16.802 values between 16.80 and 17.631 values between 17.63 and 18.461 values between 19.29 and 20.111 values between 20.11 and 20.941 values between 22.60 and 23.431 values between 23.43 and 24.261 values between 26.75 and 27.58offnegativepositiveoften

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. cup
  2. lse
  3. ��
  4. ///<
  5. .toHexString
  6. כן
  7. 实
  8. `,
  9. 轿
  10. 庾

Suppresses

  1. OMATIC
  2. unsch
  3. ""),⏎
  4. lessly
  5. -hash
  6. arine
  7. eds
  8. esh
  9. LOT
  10. -rich