Neuron Explorer

Qwen3-0.6B

'for' after praise words like 'known for'

A neuron in layer 6 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 'for' right after words that describe being famous or valued, such as 'known for', 'valued for' or 'recognized for'. Turning it up did not produce a clear change in the model's writing.

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

    soccer game.⏎It’s really too bad that we have bought into these inaccurate interpretations of this story.⏎I feel bad for the character Thomas, who is only remembered for what he apparently did not believe.⏎It’s not a stretch to say that our

  2. #2 peak

    Chemical Activity of Copper:⏎The chemical activity of copper imparts unique taste and aromatic characteristics to beverages.⏎Despite their higher cost, copper stills are valued for their ability to create high-quality and flavorful alcoholic beverages.In copper stills, a

  3. #3 peak

    outgoing, they tend to chatter a lot and they are quite active little birds. They love to interact with their "flock" and are known around the world for their exceptional talking

  4. #4 peak

    below.⏎You can read a full overview at Wikipedia here.<|endoftext|>Aloe vera has long been considered as one of Mother Nature’s miracle plants, well-regarded for its healing, antioxidant, anti-inflammatory, and moisturizing properties, among others.

  5. #5 peak

    narrative of creation. As the<|endoftext|>Cinnamon contains large amounts of very powerful polyphenolic antioxidants.⏎The cinnamon is a spice that has been appreciated since ancient times for

  6. #6 peak

    land⏎(1808–1867) Hannah was born and raised in a Quaker family in Sherwood, NY and is best recognized for her work as an abolitionist. The home she shared with her husband, S

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 design of a modern car, and the current trend in the automotive industry is towards more efficient and higher performance. Therefore, the new car should be a more efficient and higher-performance than the previous models. This is what the design should be

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

    Even though the British fleet's arrival in New York set off a flurry of panic amongst the Loyalist population. The news of the defeat was also not received well in London by King George III who is noted as saying that "after the knowledge

  2. #2 peak

    There are some tribes amongst them (for they are distinguished like the Jews by their tribes), among whom the crime of swearing by the name of the Virgin is punished with forfeiture of

  3. #3 peak

    above enactments and other allied laws for the smooth functioning of their business. Any company that doesn’t follow the rules listed in these enactments is subject to punishment by the government of India. It is to be noted that the organisations with large operations

  4. #4 peak

    welcomed spring and the dawn, why our word for "East" and "Easter" come from her name. Originally in Pagan traditions, Easter was celebrated with a feast to celebrate Ostara (the Old English spelling) the bringer of

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. L5:2203 weakly suppresses
  2. L3:2217 weakly suppresses
  3. L4:3067 weakly excites
  4. L3:1611 weakly suppresses
  5. L4:1539 weakly suppresses
  6. L5:2931 weakly suppresses
  7. L5:733 weakly excites
  8. L3:1747 weakly excites
  9. L5:2162 weakly suppresses
  10. L5:2636 weakly excites
  11. L4:1454 weakly excites
  12. L3:1651 weakly excites "on" in "went on to"
  13. L0:2779 weakly suppresses
  14. L4:2347 weakly excites
  15. L2:2938 weakly excites
  16. L2:2618 weakly suppresses
  17. L3:505 weakly excites
  18. L3:408 weakly suppresses
  19. L5:2058 weakly excites
  20. L0:1303 weakly excites
  21. L5:151 weakly excites
  22. L5:2094 weakly excites
  23. L5:2691 weakly suppresses
  24. L3:904 weakly excites

Feeds later layers

  1. L9:2451 excites 'for' giving a reason or cause
  2. L14:2404 excites
  3. L7:2581 excites
  4. L7:43 excites
  5. L18:2859 excites 'for' after an award or honour
  6. L17:1952 weakly excites
  7. L17:1820 weakly excites words of celebrating or promoting
  8. L18:2049 weakly excites
  9. L20:2795 weakly excites
  10. L17:2782 weakly suppresses 'same' in 'at the same time'
  11. L13:519 weakly excites
  12. L9:2545 weakly suppresses
  13. L13:2141 weakly excites end-of-text token after finished articles
  14. L10:1860 weakly excites
  15. L16:547 weakly excites
  16. L15:598 weakly suppresses
  17. L13:2791 weakly excites
  18. L20:1671 weakly suppresses
  19. L8:1510 weakly suppresses
  20. L20:1259 weakly excites
  21. L20:2306 weakly excites
  22. L14:882 weakly suppresses
  23. L11:374 weakly suppresses
  24. L12:2303 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 -1.92 and -1.681 values between -1.22 and -0.994 values between -0.99 and -0.7511 values between -0.75 and -0.5297 values between -0.52 and -0.291029 values between -0.29 and -0.062494 values between -0.06 and 0.18236 values between 0.18 and 0.4118 values between 0.41 and 0.644 values between 0.64 and 0.882 values between 0.88 and 1.113 values between 1.11 and 1.344 values between 1.57 and 1.811 values between 1.81 and 2.041 values between 2.04 and 2.271 values between 2.74 and 2.971 values between 7.15 and 7.39offnegativepositiveoften

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. za
  2. trib
  3. 涤
  4. virtue
  5. ertia
  6. ocrat
  7. ấm
  8. CRT
  9. Containing
  10. pedig

Suppresses

  1. anean
  2. setDefault
  3. <<(
  4. rä
  5. aneous
  6. imentary
  7. vary
  8. uale
  9. قدير
  10. bah