Neuron Explorer

Qwen3-0.6B

"is" in "there is" sentences

A neuron in layer 20 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 word "is" in phrases like "There is no evidence" or "There is a strong tradition." It goes quiet, or dips, when the sentence is a negation such as "cannot" or "not." Turning it up made the model's text drift, with no clear link to the trigger.

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

    community⏎WOODFORD COUNTY, Ky. (WKYT) – Ten years after the last resident left a historic African American community in Woodford County, there is now a new birth and a new way to remember a rich history .⏎What happened

  2. #2 peak

    without an ulceration (broken skin). There is no evidence that Stage I melanoma has spread to the lymph tissues, lymph nodes

  3. #3 peak

    & Wildlife Service (FWS).⏎The FWS announced its decision on the plant Thursday. The species remains on the FWS’s "candidate list," although there is no prediction of when, or if, it will actually receive ESA protection.⏎It

  4. #4 peak

    . There is a remarkable amount of mosaic pavement that is preserved – more than anywhere in the entire

  5. #5 peak

    . There is a thick green mat of vegetation, but little diversity to recommend to wildlife or the

  6. #6 peak

    .⏎Demonstration: This method uses an equal-arm balance rather than a domestic spring balance.⏎Class practical: A simple method for comparing the density of liquids.⏎There is a strong tradition of beginning physics with careful measurements of volumes and masses and calculations of

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: is of is Okay, I have is of is. So, is of is is of is. That's of is is of is. So, of is of is is of is. Then of is of is of is. So

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

    by itself. Therefore, a number of the more popular study questions include (but are not limited to) whether there are connections between colors and learning, whether students learn faster

  2. #2 peak

    . The deaths of Black women like Breonna Taylor, Sandra Bland, Regis Korchinski-Paquet, and others at the hands of the police cannot be forgotten. We acknowledge the pain of Black communities at large, but also recognize

  3. #3 peak

    of the coast, in consequence of which he called the peaked island to the westward of the cape, Magnetical Island: this irregularity, however, was not noticed by me in my observations near the same spot; and the difference observed by

  4. #4 peak

    ies⏎When we think about food allergies, we're much more likely to picture peanuts, shellfish or even eggs. But mild apple allergies are not uncommon. Of note, experts believe that it's the birch pollen that's

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. L17:227 suppresses
  2. L16:466 excites 'is' after a long subject phrase
  3. L4:16 excites
  4. L14:924 excites
  5. L19:576 suppresses
  6. L4:4 weakly suppresses
  7. L15:599 weakly excites
  8. L18:93 weakly suppresses
  9. L15:286 weakly excites
  10. L19:1997 weakly excites
  11. L14:149 weakly suppresses
  12. L18:1389 weakly suppresses
  13. L5:351 weakly excites
  14. L19:1727 weakly excites
  15. L15:681 weakly suppresses
  16. L17:137 weakly excites
  17. L14:453 weakly suppresses
  18. L1:1409 weakly excites
  19. L11:1099 weakly suppresses
  20. L15:629 weakly excites
  21. L11:31 weakly suppresses
  22. L16:1168 weakly excites 'to' after verbs of change
  23. L17:327 weakly suppresses
  24. L17:706 weakly excites

Feeds later layers

  1. L22:228 suppresses
  2. L23:127 excites
  3. L26:53 excites
  4. L21:2478 weakly suppresses
  5. L25:9 weakly excites
  6. L22:666 weakly suppresses
  7. L23:1013 weakly suppresses
  8. L24:100 weakly excites
  9. L22:353 weakly suppresses
  10. L24:2880 weakly suppresses
  11. L26:1257 weakly suppresses
  12. L22:1434 weakly suppresses
  13. L22:1138 weakly suppresses
  14. L22:2821 weakly suppresses
  15. L21:2981 weakly suppresses
  16. L27:604 weakly suppresses
  17. L23:128 weakly excites 'for' as in 'for example'
  18. L25:247 weakly excites
  19. L22:2769 weakly suppresses
  20. L24:845 weakly suppresses
  21. L27:382 weakly suppresses
  22. L24:2684 weakly suppresses
  23. L23:30 weakly excites
  24. L21:930 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.

1 values between -16.56 and -15.571 values between -13.58 and -12.592 values between -12.59 and -11.603 values between -10.60 and -9.616 values between -9.61 and -8.629 values between -8.62 and -7.6213 values between -7.62 and -6.6315 values between -6.63 and -5.6431 values between -5.64 and -4.6456 values between -4.64 and -3.6596 values between -3.65 and -2.66220 values between -2.66 and -1.66476 values between -1.66 and -0.671879 values between -0.67 and 0.33547 values between 0.33 and 1.32117 values between 1.32 and 2.3187 values between 2.31 and 3.3171 values between 3.31 and 4.3047 values between 4.30 and 5.2935 values between 5.29 and 6.2936 values between 6.29 and 7.2822 values between 7.28 and 8.2726 values between 8.27 and 9.2732 values between 9.27 and 10.2613 values between 10.26 and 11.2517 values between 11.25 and 12.2512 values between 12.25 and 13.246 values between 13.24 and 14.238 values between 14.23 and 15.237 values between 15.23 and 16.222 values between 16.22 and 17.213 values between 17.21 and 18.214 values between 18.21 and 19.202 values between 19.20 and 20.193 values between 20.19 and 21.192 values between 21.19 and 22.181 values between 22.18 and 23.17offnegativepositiveoften

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. considered
  2. taşı
  3. dışı
  4. 她说
  5. 实例
  6. seen
  7. particularly
  8. several
  9. LinkId
  10. some

Suppresses

  1. ivity
  2. 性
  3. iale
  4. edImage
  5. 性和
  6. 起了
  7. ed
  8. edy
  9. ivities
  10. imony