Neuron Explorer

Qwen3-0.6B

'in' after a noun or participle

A neuron in layer 16 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 'in' (or 'into') when it follows a word like 'present', 'contained', 'hidden' or 'laden' and names where something is located or contained. When it was turned up, the written text drifted into unrelated topics, so the push did not clearly bring out this idea.

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

    a rule regarding the sale and distribution of flavor-infused vaping cartridges. Parents need to be aware of the health risks associated with vaping, and share information about chemicals in vaping products with their children.⏎Because vaping devices can be easily concealed and are small

  2. #2 peak

    at Stanford University and the prestigious Cyclotron Road program at Lawrence Berkeley National Lab, recently announced the advancement of a new electrochemical technology that converts the carbon dioxide content in raw bi

  3. #3 peak

    foods that contain gluten. Gluten is a protein present in a certain number of cereals: wheat, rye, oats and barley.

  4. #4 peak

    ist hate speech in the print and electronic media”.⏎The UN committee asked the Australian Government to report back to it in detail in the<|endoftext|>Discover the words hidden in the grid, written through one another in eight horizontal, vertical and diagonal directions.

  5. #5 peak

    .⏎Foods with added probiotics have are also recently available:-⏎– like probiotic chocolate, snack bars and teabags⏎– the probiotic bacteria in these foods is often freeze dried⏎– these foods aren’t fermented, but they

  6. #6 peak

    of mesothelioma in the Iron Range miners.⏎The textile industry used asbestos in heat resistant products like heat–resistant pads, asbestos gloves, and woven into cloth coverings and protective blankets.⏎There was extensive use of asbestos in brake pads

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: my family of the family of my home. I think of my family of my school. my school. my school. my university. my university. my college. my college. my college. my college. my university. my university.

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

    SON MANUAL (1993)||⏎previous taxon |⏎Jepson Interchange (more information)⏎©Copyright 1993 by the Regents of the University of California⏎AND IS MAINTAINED FOR ARCHIVAL PURPOSES ONLY

  2. #2 peak

    of E. canguzua.⏎• Marcus, Er. (1955) Opisthobranchia from Brazil. Boletim da Fac

  3. #3 peak

    liberty of contract” in a 5-4 decision.⏎Justice Oliver Wendell Holmes Jr, dissenter in the case, and Roscoe Pound (Dean of Harvard Law School), expressed views around the time that were termed “sociological

  4. #4 peak

    Volcanism Program, 2007. Report on Macdonald (Undersea Features) (Wunderman, R., ed.). Bulletin of the Global Volcanism Network, 32:1. Smithsonian Institution. https

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. L15:871 suppresses
  2. L15:1464 suppresses
  3. L12:2833 excites
  4. L15:632 excites
  5. L14:1728 suppresses "of" after a numbered section or title
  6. L11:2207 suppresses "of" after a noun in trademark and title phrases
  7. L12:2807 suppresses
  8. L12:1040 excites
  9. L12:1904 suppresses
  10. L12:2756 excites "of" after a noun of quantity or container
  11. L11:1174 suppresses
  12. L15:2361 excites
  13. L9:2862 suppresses
  14. L12:2889 suppresses
  15. L15:2330 excites
  16. L9:2219 suppresses
  17. L9:2545 suppresses
  18. L11:2037 excites
  19. L8:3068 suppresses
  20. L15:3006 excites
  21. L15:942 suppresses
  22. L9:1909 excites
  23. L14:2512 excites
  24. L9:1595 excites

Feeds later layers

  1. L19:1733 suppresses
  2. L17:2405 excites
  3. L17:1574 suppresses
  4. L17:2733 excites
  5. L24:490 excites
  6. L17:2906 suppresses
  7. L18:1674 suppresses
  8. L17:1045 excites
  9. L19:2937 suppresses unclear: mixed possessives and fragments
  10. L27:886 suppresses
  11. L21:2481 suppresses
  12. L18:2472 suppresses
  13. L17:2314 suppresses
  14. L19:1354 suppresses
  15. L18:604 excites
  16. L21:2075 suppresses
  17. L21:933 excites
  18. L21:2846 suppresses 'at' after a job title, before an institution
  19. L18:2090 suppresses
  20. L17:1363 excites
  21. L18:158 suppresses
  22. L17:932 suppresses
  23. L17:978 suppresses
  24. L17:2199 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 -5.25 and -4.991 values between -4.47 and -4.212 values between -3.69 and -3.444 values between -3.44 and -3.182 values between -3.18 and -2.922 values between -2.92 and -2.662 values between -2.66 and -2.406 values between -2.40 and -2.148 values between -2.14 and -1.8816 values between -1.88 and -1.6216 values between -1.62 and -1.3624 values between -1.36 and -1.1148 values between -1.11 and -0.8572 values between -0.85 and -0.59145 values between -0.59 and -0.33703 values between -0.33 and -0.072335 values between -0.07 and 0.19338 values between 0.19 and 0.4568 values between 0.45 and 0.7139 values between 0.71 and 0.9723 values between 0.97 and 1.2212 values between 1.22 and 1.4810 values between 1.48 and 1.749 values between 1.74 and 2.004 values between 2.00 and 2.263 values between 2.26 and 2.523 values between 2.52 and 2.782 values between 2.78 and 3.043 values between 3.04 and 3.292 values between 3.29 and 3.551 values between 3.55 and 3.811 values between 3.81 and 4.071 values between 4.07 and 4.332 values between 4.85 and 5.11offnegativepositiveoften

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. 男孩子
  2. ables
  3. 建筑材料
  4. 什么东西
  5. şı
  6. ucus
  7. 再见
  8. nextProps
  9. reeze
  10. 汉子

Suppresses

  1. courts
  2. gium
  3. {'
  4. 加工厂
  5. {"
  6. Northern
  7. museums
  8. crib
  9. 含
  10. warehouses