Neuron Explorer

Qwen3-0.6B

'there' before are or is

A neuron in layer 22 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 word 'there' when it starts a sentence like 'there are many' or 'there is evidence', especially after words like 'However' or 'Currently'. Turning it up made the model repeat forms of 'is' and 'were' over and over, which fits its role around these verbs.

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

    liver problems, he may start suffering from a metabolic disorder, liver disease, or type 2 diabetes. However, there are multiple drinks and foods that are good for liver repair and maintenance.⏎Read this

  2. #2 peak

    four basic categories. These include wood, dog, deer, and spinose ear ticks. However, there are many specific tick species within each of these categories. Some ticks are found across

  3. #3 peak

    time. Additionally, smart work can lead to innovation and creativity, as individuals are encouraged to think outside the box and find new solutions to old problems.⏎However, there are also disadvantages to smart work. First, it can lead to a lack of

  4. #4 peak

    feet, as well as issues with their reproductive systems. The virus does not affect humans and it belongs to the Picornaviridae family. Currently, there are seven known strains, of which only A and C have been verified in the

  5. #5 peak

    and a sense of purpose. Additionally, hard work can often lead to greater mastery of skills, as repetition and practice can lead to improvement over time.⏎However, there are also several disadvantages to hard work. First, it can be time-consuming,

  6. #6 peak

    . It is even accepted that the face of the Sphinx bears the likeness of Khafre (Chephren), the Pharaoh during that time.⏎However, there was no written record from that era regarding the Sphinx or its construction. In addition

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 process of the process of the process of the process of the process... so that you can make a decision as your own decision, its the process of the process of the process of the process... that you as your own decision, the

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

    of the most frustrating behaviors a dog owner can experience. It not only disturbs your sleep but also gives you the sense that the dog is not comfortable. So what are the reasons why dogs growl?⏎Here’s The Sort Answer To Why Your

  2. #2 peak

    of ink and carbon paper. A significant step forward towards the modern-day typewriter was achieved by the Italian lawyer Giuseppe Ravizza, whose “cembalo scrivano” (scribe harpsichord) had keys resembling

  3. #3 peak

    Without worrying about the problems, he started removing water from the sea with his beak.⏎Seeing this, another bird sitting beside him said, “What are you doing, you’ll not get any benefit from this.”⏎The crow replied

  4. #4 peak

    of gems, known only to kings.” Pliny described six varieties, among which the Indian, having six pointed angles, and also resembling two pyramids (turbines, whip-tops) placed base to

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. L20:1729 strongly excites
  2. L17:1618 excites
  3. L21:2536 excites
  4. L18:427 excites
  5. L21:1362 excites
  6. L17:3071 excites
  7. L19:2942 excites
  8. L20:2622 suppresses
  9. L20:2222 suppresses
  10. L15:1007 suppresses
  11. L20:2508 weakly suppresses
  12. L16:2500 weakly suppresses
  13. L19:572 weakly excites
  14. L21:618 weakly suppresses
  15. L20:2360 weakly excites
  16. L21:1622 weakly excites
  17. L20:2994 weakly excites
  18. L19:482 weakly suppresses
  19. L20:2335 weakly excites comma ending an 'if' clause
  20. L21:2808 weakly excites
  21. L5:1094 weakly suppresses
  22. L17:2162 weakly suppresses
  23. L18:1050 weakly suppresses
  24. L17:1480 weakly excites

Feeds later layers

  1. L26:109 suppresses
  2. L24:2880 excites
  3. L24:299 excites
  4. L27:650 excites
  5. L27:500 excites
  6. L24:319 excites
  7. L27:804 excites
  8. L25:9 suppresses
  9. L23:1013 excites
  10. L24:565 excites
  11. L23:220 excites
  12. L25:822 excites
  13. L26:2947 excites
  14. L26:1097 excites
  15. L25:1502 weakly excites
  16. L26:440 weakly excites
  17. L26:1182 weakly excites
  18. L25:698 weakly excites
  19. L25:613 weakly excites
  20. L23:1572 weakly excites
  21. L23:1264 weakly suppresses
  22. L24:3019 weakly suppresses
  23. L24:1811 weakly excites
  24. L24:1423 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 -13.71 and -8.745 values between -8.74 and -3.763798 values between -3.76 and 1.2185 values between 1.21 and 6.197 values between 6.19 and 11.163 values between 11.16 and 16.132 values between 16.13 and 21.113 values between 21.11 and 26.081 values between 31.06 and 36.031 values between 65.87 and 70.851 values between 135.51 and 140.481 values between 180.28 and 185.25offnegativepositiveoften

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. are
  2. is
  3. ARE
  4. are
  5. were
  6. _are
  7. Are
  8. 的是
  9. Are
  10. .are

Suppresses

  1. YYY
  2. ahoo
  3. _Abstract
  4. ^.
  5. 好吗
  6. ackle
  7. 治好
  8. enment
  9. 平均水平
  10. ixo