Neuron Explorer

Qwen3-0.6B

full stop ending a citation title

A neuron in layer 11 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

The neuron fires on the period that comes right after the title of a book or article in a reference list, such as 'Philosophy of Physics.' before the publisher. Turning it up gave off-topic, repetitive text, so the steering doesn't reveal much more.

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

    , J. H., & Talaga, M. C. (2017). Handbook of clinical psychopharmacology for therapists (8th ed.). Oakland, CA: New Harbinger.⏎Chapter 8, “Bip

  2. #2 peak

    Revising to create “showing” sentences systematically helps improve writing competence.⏎Dyson, A. (2008). Playing with textual toys. In Flood, J., Heath, S.B., & Lapp, D.

  3. #3 peak

    |Through your library||Configure|⏎Similar books and articles⏎Peter Kosso (1998). Appearance and Reality: An Introduction to the Philosophy of Physics. Oxford University Press.⏎Karl Rogers (2005). On the Met

  4. #4 peak

    .⏎Lawrence Sklar (1992). Philosophy of Physics. Westview Press.⏎Edward G. Ruestow (1973).

  5. #5 peak

    fix it)|⏎|Through your library||Configure|⏎Similar books and articles⏎James Behuniak Jr (2011). Naturalizing Mencius. Philosophy East and West 61 (3):492-51

  6. #6 peak

    or not alcohol<|endoftext|>Using groundwater for hydro electric power generation⏎Bardsley, W.E. (1995). Using groundwater for hydro electric power generation. Journal of Hydrology (NZ), 34(1), 1-

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: "A” is the link of the form: https://www.google.com/search?q=exact:msi.dla.arc.arc.arc.arc.arc.arc.au.au.au.au.au.au.au.au.au.au.au.au.au.au.au.au.au.au

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

    child’s healthcare provider about glucagon emergency kits for home and school.<|endoftext|>|This article needs additional citations for verification. (July 2012) (Learn how and when to remove this template message)|⏎Sendust is a magnetic

  2. #2 peak

    Pakistan⏎||It has been suggested that this article be merged with Blasphemy law#Pakistan. (Discuss) Proposed since March 2013.|⏎The Pakistan Penal Code

  3. #3 peak

    never came to fruition, the building plots which would be home Carlisle Memorial Church and the Belfast Orange Hall.<|endoftext|>|This article needs additional citations for verification. (March 2010)|⏎|Transformer with two windings and iron

  4. #4 peak

    , giving support for the lateral method of adaptation.<|endoftext|>||This article includes a list of references, but its sources remain unclear because it has insufficient inline citations. (May 2009)|⏎Lexicography is divided into two separate

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. L2:2791 weakly suppresses
  2. L9:1050 weakly excites
  3. L9:1383 weakly suppresses
  4. L1:2981 weakly suppresses
  5. L8:172 weakly suppresses
  6. L1:200 weakly suppresses
  7. L2:2057 weakly excites
  8. L5:515 weakly suppresses
  9. L5:28 weakly excites
  10. L2:2879 weakly excites
  11. L10:2636 weakly suppresses
  12. L7:2509 weakly suppresses
  13. L7:1782 weakly excites
  14. L3:488 weakly excites
  15. L1:897 weakly suppresses
  16. L6:2859 weakly excites
  17. L5:1542 weakly suppresses
  18. L10:2758 weakly excites
  19. L4:1332 weakly suppresses
  20. L5:1052 weakly excites
  21. L8:2168 weakly excites
  22. L9:479 weakly excites
  23. L9:2387 weakly suppresses
  24. L8:1571 weakly excites

Feeds later layers

  1. L13:899 excites
  2. L16:808 weakly excites
  3. L16:2880 weakly excites
  4. L22:791 weakly excites
  5. L12:1854 weakly excites
  6. L15:1955 weakly suppresses
  7. L16:1364 weakly excites
  8. L17:905 weakly suppresses
  9. L25:2068 weakly suppresses
  10. L20:852 weakly suppresses
  11. L15:2536 weakly excites
  12. L22:1655 weakly excites
  13. L13:1112 weakly excites
  14. L18:1987 weakly excites
  15. L26:2230 weakly excites
  16. L16:2922 weakly suppresses
  17. L15:1078 weakly suppresses
  18. L14:1239 weakly excites
  19. L15:1425 weakly excites
  20. L21:397 weakly suppresses
  21. L22:2996 weakly excites
  22. L23:1003 weakly excites
  23. L18:2346 weakly excites
  24. L16:2216 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 -0.95 and -0.861 values between -0.86 and -0.781 values between -0.78 and -0.704 values between -0.70 and -0.613 values between -0.61 and -0.5322 values between -0.53 and -0.4552 values between -0.45 and -0.37161 values between -0.37 and -0.28262 values between -0.28 and -0.20556 values between -0.20 and -0.12847 values between -0.12 and -0.04861 values between -0.04 and 0.05585 values between 0.05 and 0.13313 values between 0.13 and 0.21140 values between 0.21 and 0.3061 values between 0.30 and 0.3823 values between 0.38 and 0.465 values between 0.46 and 0.546 values between 0.54 and 0.632 values between 0.63 and 0.711 values between 1.54 and 1.621 values between 2.28 and 2.36offnegativepositiveoften

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. semantics
  2. dispos
  3. lek
  4. нул
  5. ContextHolder
  6. &id
  7. zik
  8. 东方财富
  9. &B
  10. ажд

Suppresses

  1. 首轮
  2. SEARCH
  3. cribed
  4. quali
  5. orca
  6. Pré
  7. 卵
  8. وق
  9. EAR
  10. HMS