Neuron Explorer

Qwen3-0.6B

words that expect an of, to or for next

A neuron in layer 12 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on a word in an informative sentence that is usually followed by 'of', 'to' or 'for', like 'symbol' in 'a symbol of' or 'potential' in 'potential to'. Turning it up did not produce a clear visible theme in the model's continuations.

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

    it has spread to almost every state. With a widespread, shallow root system, it is ideal for thin soils where it will prevent soil erosion; it is often used for strip-mine reclamation projects barren due to mining debris and acid soils.⏎

  2. #2 peak

    Kreimer et al., 2011). More information on HPV and cervical cancer.⏎In 2012 the U.S. Preventive Services Task Force and a coalition of health organizations published new guidelines for cervical cancer screening.

  3. #3 peak

    Definition of endoplasm⏎n. - The protoplasm in the interior of a cell. 2⏎The

  4. #4 peak

    what must happen in Advent, what Advent must be, if we want to live.”⏎The candle is an image of the Christian calling, and is also a symbol of the witness Father Delp gave in his suffering and death. In a sermon

  5. #5 peak

    . "Fundamentally this question must be answered to advance the knowledge base and to know all the risks and benefits of cancer treatment. This research has the potential to reach across numerous scientific disciplines, and may one day improve the lives of patients worldwide

  6. #6 peak

    EDX). This technique can provide not only information on the shape of aerosol particles, but also spatial relationships between the compounds that comprise those particles.⏎The power of optical microscopy is multiplied when coupled to techniques that can chemically speciate the object

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: I think this is a way of thinking. I think this is a way of thinking. I think I have been thinking about lately. I think this is a way of thinking. I think I have been thinking about lately. I think I

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

    baths. The antechamber leads to a courtyard with columns and walkways. The antechamber is not directly lined up with the axis of the courtyard – intentional? Diverts public traffic to the right side – which is a shorter pathway.

  2. #2 peak

    classroom, when the class discusses and analyzes a work of fiction, poetry, or creative non-fiction, the main business of that discussion is to determine why the author made the

  3. #3 peak

    to email@example.com.⏎Why are we doing this?⏎The purpose of making all information about children's rights available in one place is to build a clearer picture of some of the repeated violations of children's rights in

  4. #4 peak

    Lutz: Tell me, why do you insist on referring to yourself in the third person? It's intensely irritating!⏎- Hercule Poirot: It helps Poirot to keep a distance from his genius.⏎

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. L10:2075 excites
  2. L10:1741 excites
  3. L8:1890 suppresses
  4. L7:3068 suppresses
  5. L10:1977 excites
  6. L11:2425 suppresses
  7. L7:1871 suppresses
  8. L6:1737 suppresses
  9. L8:2012 excites
  10. L10:1427 weakly suppresses
  11. L7:530 weakly suppresses
  12. L9:2119 weakly excites
  13. L9:2008 weakly suppresses
  14. L8:2016 weakly excites
  15. L9:2543 weakly suppresses
  16. L11:1721 weakly suppresses
  17. L11:3070 weakly suppresses
  18. L11:177 weakly suppresses
  19. L8:1771 weakly suppresses
  20. L10:2790 weakly excites
  21. L7:2535 weakly suppresses
  22. L0:1051 weakly suppresses
  23. L10:278 weakly suppresses
  24. L8:2544 weakly suppresses

Feeds later layers

  1. L16:2340 suppresses
  2. L17:2593 suppresses
  3. L21:2060 suppresses
  4. L13:481 suppresses 'because' introducing a reason
  5. L15:781 suppresses
  6. L14:1666 suppresses
  7. L14:2244 suppresses
  8. L14:852 suppresses
  9. L16:2685 suppresses
  10. L13:2172 suppresses
  11. L13:2414 suppresses
  12. L14:2891 suppresses
  13. L17:2517 suppresses
  14. L17:405 excites
  15. L13:868 excites
  16. L14:296 excites
  17. L13:320 excites
  18. L17:2898 weakly suppresses
  19. L15:2494 weakly suppresses
  20. L16:1510 weakly suppresses
  21. L20:1861 weakly suppresses
  22. L18:923 weakly suppresses
  23. L14:927 weakly suppresses
  24. L13:1418 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.

2 values between -3.40 and -3.281 values between -2.91 and -2.781 values between -2.78 and -2.661 values between -2.66 and -2.531 values between -2.41 and -2.281 values between -2.28 and -2.162 values between -2.16 and -2.031 values between -1.91 and -1.782 values between -1.78 and -1.662 values between -1.66 and -1.544 values between -1.54 and -1.411 values between -1.29 and -1.164 values between -1.16 and -1.044 values between -1.04 and -0.917 values between -0.91 and -0.7919 values between -0.79 and -0.6621 values between -0.66 and -0.5434 values between -0.54 and -0.4172 values between -0.41 and -0.29240 values between -0.29 and -0.16826 values between -0.16 and -0.042007 values between -0.04 and 0.09475 values between 0.09 and 0.21102 values between 0.21 and 0.3339 values between 0.33 and 0.4613 values between 0.46 and 0.5810 values between 0.58 and 0.714 values between 0.71 and 0.836 values between 0.83 and 0.963 values between 0.96 and 1.081 values between 1.08 and 1.211 values between 1.21 and 1.331 values between 1.46 and 1.58offnegativepositiveoften

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. ansen
  2. inkle
  3. 发生
  4. -shot
  5. adders
  6. Even
  7. lem
  8. even
  9. rozen
  10. asonry

Suppresses

  1. 的目的
  2. 是为了
  3. 目的是
  4. managedType
  5. wear
  6. ativas
  7. ует
  8. 之意
  9. irse
  10. purpose