Neuron Explorer

Qwen3-0.6B

'in' in 'in progress' phrases

A neuron in layer 25 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 small word 'in' in set phrases such as 'work in progress', 'in stock', 'in place' and 'in the pipeline'. When it was turned up, the model kept repeating the word 'progress', which fits what it pushes toward.

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

    technology, the electric motor is a popular drive train alternative in electric vehicles. The development of the electric motor as a drive train for electric vehicles is still a work in progress however. The first genuine electric motor was produced as early as 18

  2. #2 peak

    contains information for science fairs and reproducibles (some in color).⏎Vendor: BJU Press⏎Publication Date: 2010⏎Availability: In Stock⏎Series: BJU Press Science⏎Be the first to write a review

  3. #3 peak

    14 July 2013).⏎This user's manual is a work in progress. We shall supplement it as we use the Actoz°° system created

  4. #4 peak

    what sources?⏎- What makes a duplicate a duplicate?⏎- Should records be merged? If so, how is a match determined?⏎- What validation rules are currently in place? Do these help or hinder?⏎Likewise, take advantage of Field Trip

  5. #5 peak

    in the inner Bulgarian regions. Currently, an active surveillance system is in place for dairy factory employees and persons considered at risk after outbreaks in ruminants

  6. #6 peak

    avrovo national park, after sustained environmental criticism centred on the potential extinction of the Balkan lynx.⏎But other projects are still in the pipeline, even if much of the energy they produce is destined for export.⏎

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 saw a test from the local school for the beginning of the school year on the web. The test had on one of its had on one of its had on one of its had on one of its had on one of its had on

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

    Defense Advanced Research Projects Agency and the U.S. Army Research Office.<|endoftext|>2011 State Poverty Data Underscore the Need to Protect Programs for Low-Income Women⏎New data from the Census was just released, and NWLC’s

  2. #2 peak

    by children calculate their GDA label on the dietary reference values for adults, for instance, which leads to excessive consumption of critical nutrients such as added sugar, said INSP.⏎Interpreting the information presented in the GDA format to understand a

  3. #3 peak

    ) that can be used to support a particular viewpoint.⏎So, what to do about it?⏎To see clear evidence ignored, distorted or diluted in favour of ill-informed subjective views leads

  4. #4 peak

    meaning the word contributes to the passage in which it appears.”⏎For more information, The report card can be found online: www.nationsreportcard.gov.MORE IN Vermont News⏎- Most Popular⏎- Most Emailed⏎- MEDIA G

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. L22:535 excites
  2. L20:1574 excites
  3. L4:289 excites
  4. L17:706 excites
  5. L21:1935 suppresses
  6. L24:8 suppresses
  7. L16:1557 excites 'in' meaning 'in this case' or 'in between'
  8. L22:530 weakly suppresses
  9. L19:704 weakly excites
  10. L3:2746 weakly excites
  11. L3:493 weakly suppresses
  12. L20:1228 weakly excites
  13. L16:1807 weakly suppresses 'in' after a motion verb
  14. L14:1625 weakly excites the word 'in' after a noun
  15. L19:1351 weakly suppresses
  16. L23:1911 weakly suppresses
  17. L0:1882 weakly excites unclear: 'high tech' and mid-word pieces
  18. L22:151 weakly suppresses
  19. L7:10 weakly suppresses
  20. L4:2 weakly suppresses
  21. L20:1120 weakly excites
  22. L6:13 weakly excites
  23. L0:781 weakly suppresses
  24. L17:2131 weakly excites

Feeds later layers

  1. L27:2701 weakly excites
  2. L26:2165 weakly excites
  3. L27:814 weakly excites
  4. L26:170 weakly excites
  5. L27:2147 weakly suppresses
  6. L26:1269 weakly suppresses
  7. L26:893 weakly suppresses
  8. L26:1710 weakly suppresses
  9. L27:2307 weakly excites
  10. L26:894 weakly suppresses
  11. L27:1263 weakly excites
  12. L27:2149 weakly excites
  13. L27:1341 weakly suppresses
  14. L26:1131 weakly suppresses
  15. L27:1121 weakly excites
  16. L26:1223 weakly suppresses
  17. L27:596 weakly suppresses
  18. L27:1597 weakly excites
  19. L26:1627 weakly excites
  20. L27:2034 weakly suppresses
  21. L26:2421 weakly suppresses
  22. L26:2857 weakly excites
  23. L27:2182 weakly suppresses
  24. L27:591 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 -21.52 and -20.051 values between -15.63 and -14.161 values between -14.16 and -12.691 values between -12.69 and -11.221 values between -8.28 and -6.812 values between -6.81 and -5.345 values between -5.34 and -3.873 values between -3.87 and -2.4044 values between -2.40 and -0.933685 values between -0.93 and 0.55106 values between 0.55 and 2.027 values between 2.02 and 3.4911 values between 3.49 and 4.968 values between 4.96 and 6.438 values between 6.43 and 7.903 values between 7.90 and 9.374 values between 9.37 and 10.844 values between 10.84 and 12.311 values between 12.31 and 13.785 values between 13.78 and 15.252 values between 15.25 and 16.723 values between 18.19 and 19.662 values between 35.84 and 37.31offnegativepositiveoften

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. progress
  2. ProcessEvent
  3. INES
  4. 行动
  5. 进步
  6. ilig
  7. action
  8. progress
  9. _PROGRESS
  10. Progress

Suppresses

  1. Dar
  2. OM
  3. htdocs
  4. gren
  5. geben
  6. entering
  7. ocity
  8. juana
  9. cretion
  10. formed