Neuron Explorer

Qwen3-0.6B

comma before a clause that contrasts a claim

A neuron in layer 21 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 commas and quote marks that come right after a claim, where the sentence is about to add a contrast or an explanation. When turned up, the model got stuck repeating the word 'when'. Turned down, it repeated 'though'.

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

    in the concentration of species in solution.⏎It’s frustrating that the article makes it seem like the applications of a really awesome discovery in pure science are mere moments away, rather than explaining why a discovery like this is awesome and encouraging peoples’ interest in

  2. #2 peak

    species as a group often has been exaggerated and misrepresented. For example, ecologists and conservationists often describe non-native species as the world’s second greatest extinction threat, despite the fact that existing data shows this clearly not to be the case. It

  3. #3 peak

    tendency, in some quarters, to jump to a solution like bariatric surgery, and rhapsodize about what an acceptable safe effective treatment alternative it is, without ever examining the underlying reasons for the obesity.⏎Reasons for obesity⏎Probably

  4. #4 peak

    tips for self-care for a healthy, active life.⏎1. Don’t lose weight⏎Many people spend a lifetime trying to restrict their eating to be ‘thin’, and it’s no wonder when this is so often portrayed as the height of fashion

  5. #5 peak

    probably noticed that many executives – and customers – don’t understand the difference between these terms. They most likely go as far as referring to them as the same processes, which they’re definitely not. Let’s figure out

  6. #6 peak

    and not serious.⏎Many people think they are constipated when, in fact, their bowel movements are regular. For example, some people believe they are constipated, or irregular, if they do not have a bowel movement every day. However,

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: though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though though

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

    ---⏎Search for questions⏎Still looking for answers? Try searching for what you seek or ask your own question.⏎Posted 31 Aug 2009 • 1 answer⏎Posted 20 Jan 2010 •

  2. #2 peak

    most common. In countries where only monogamous marriage is legally valid, de facto polygamy is generally permitted, as long as adultery is not illegal. In areas like these, where polyg

  3. #3 peak

    without any cross-task interference.⏎Its modular nature affords extensive customisation. The major functional components such as the number of threads, caches and DSP resources are optional, giving the SoC designer the opportunity to influence performance, power and die size to

  4. #4 peak

    used depends on the condition being treated. Generally, most practitioners recommend 5 to 15 grams of dried chrysanthemum, usually in an infusion or decoction. Some practitioners recommend a type of chrys

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. L14:35 suppresses
  2. L11:223 suppresses
  3. L18:366 excites comma inside numbers and dates
  4. L13:1219 weakly suppresses
  5. L20:750 weakly suppresses unclear: chinese academic text, 'when' loops
  6. L12:158 weakly excites
  7. L20:380 weakly excites
  8. L7:471 weakly suppresses
  9. L15:184 weakly suppresses
  10. L9:1128 weakly suppresses
  11. L5:1465 weakly suppresses
  12. L2:740 weakly excites
  13. L11:2745 weakly suppresses
  14. L14:2525 weakly excites
  15. L12:325 weakly suppresses
  16. L16:441 weakly excites
  17. L10:556 weakly suppresses
  18. L18:409 weakly suppresses
  19. L17:118 weakly excites comma after a clause ending
  20. L19:1983 weakly excites
  21. L17:2541 weakly suppresses comma before but also in not only constructions
  22. L6:2407 weakly suppresses
  23. L20:49 weakly suppresses
  24. L18:646 weakly excites

Feeds later layers

  1. L25:633 strongly excites
  2. L23:1645 excites
  3. L27:659 excites
  4. L23:449 excites
  5. L22:915 weakly suppresses digits inside historical years
  6. L23:677 weakly suppresses
  7. L27:1419 weakly excites
  8. L24:876 weakly suppresses
  9. L23:728 weakly suppresses
  10. L26:1199 weakly suppresses
  11. L25:1997 weakly excites
  12. L26:388 weakly suppresses
  13. L24:3055 weakly excites
  14. L27:1084 weakly suppresses
  15. L27:2352 weakly suppresses
  16. L22:2604 weakly suppresses
  17. L25:1593 weakly excites
  18. L23:2814 weakly excites
  19. L23:793 weakly excites
  20. L22:636 weakly excites
  21. L23:1157 weakly excites
  22. L26:850 weakly excites
  23. L22:1958 weakly suppresses
  24. L27:2168 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 -17.64 and -17.011 values between -11.99 and -11.361 values between -8.85 and -8.233 values between -7.60 and -6.974 values between -6.97 and -6.343 values between -6.34 and -5.721 values between -5.72 and -5.092 values between -5.09 and -4.463 values between -4.46 and -3.835 values between -3.83 and -3.2112 values between -3.21 and -2.5811 values between -2.58 and -1.9522 values between -1.95 and -1.32192 values between -1.32 and -0.691881 values between -0.69 and -0.071638 values between -0.07 and 0.5673 values between 0.56 and 1.1918 values between 1.19 and 1.8210 values between 1.82 and 2.445 values between 2.44 and 3.0711 values between 3.07 and 3.705 values between 3.70 and 4.331 values between 4.33 and 4.954 values between 4.95 and 5.581 values between 6.84 and 7.47offnegativepositiveoften

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. when
  2. when
  3. 时
  4. 的时候
  5. When
  6. When
  7. cuando
  8. 当
  9. →when
  10. quando

Suppresses

  1. though
  2. though
  3. although
  4. Though
  5. Though
  6. 一回事
  7. although
  8. aunque
  9. Although
  10. 虽然