Neuron Explorer

Qwen3-0.6B

middle of split technical words

A neuron in layer 8 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 the middle pieces of longer technical or scientific words that are split into several parts, such as 'gr-ating', 'R-aman' and 'Pale-olithic'. It seems to track word fragments in specialist vocabulary. Turning it up made the output change topic without a clear pattern.

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

    perm Phylogeny Website (Family: Loranthaceae, Order: Santalales)⏎Text by A. L. Quirico⏎| ||Key to the genera|| |⏎|1||Terrestrial shrubs, parasitic upon the

  2. #2 peak

    essentially the same for light from the distant nebulae as in light from laboratory sources. The assumption has been tested by measuring red-shifts both with a grating spectrograph and a prism spectrograph. The fact that the shifts are the

  3. #3 peak

    influence of aerosols on the global climate.⏎Ault, A.P., et al. “Raman Microspectroscopy and Vibrational Sum Frequency Generation Spectroscopy as Probes

  4. #4 peak

    backhoe trenches, and seven⏎excavation units. Two buried horizons were located on 48TE414. A Late⏎Prehistoric point and a

  5. #5 peak

    between the compounds that comprise those particles.⏎The power of optical microscopy is multiplied when coupled to techniques that can chemically speciate the object being magnified. Raman microspectroscopy reveals information about the composition of single aerosol particles by probing

  6. #6 peak

    forms a mass of stalactites that gradually fills in the base of the projecting terracette.⏎Collapse features are fairly common on most of the Mammoth terraces. Water flowing through near-surface channels weakens the overlying travertine

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 idea of a concept being put into practice, but it is just a thought in the memory (in the form of a procedure) for the following instances, and then it gets into the database of the database of the database of the database

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

    turn innovative ideas into finished products, often using 3D printing. As such, Fracassi

  2. #2 peak

    made statements about how the National Research Council isn’t going to pursue projects with immediate commercial applications and it makes me sad. Little curiosities like this can grow into amazing large-scale processes with a bit of support, but a lack of funding threatens

  3. #3 peak

    virtue. No amount of scolding, criticism, and punishment can turn a regrettable situation around. When you are not chided and you deserved to

  4. #4 peak

    is scientifically illustrating the educational power of things like play, sports, music, memorization and reflection. What nourishes the human spirit, the whole person, it turns out, also hones our minds.⏎An improvisational storytelling class of 5

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. L6:2821 weakly suppresses
  2. L7:2332 weakly excites
  3. L4:2585 weakly suppresses
  4. L2:2132 weakly excites
  5. L7:2991 weakly suppresses
  6. L4:2179 weakly excites
  7. L0:2296 weakly suppresses
  8. L6:3018 weakly excites
  9. L7:3036 weakly excites
  10. L7:1018 weakly suppresses
  11. L4:2444 weakly suppresses
  12. L4:3042 weakly suppresses
  13. L7:2499 weakly excites
  14. L4:1149 weakly excites
  15. L7:2186 weakly suppresses
  16. L5:2933 weakly excites
  17. L6:579 weakly excites
  18. L1:1403 weakly suppresses
  19. L3:1930 weakly excites
  20. L7:2452 weakly suppresses
  21. L7:2474 weakly suppresses
  22. L1:2675 weakly excites
  23. L3:2181 weakly suppresses comma inside numbers
  24. L6:2720 weakly excites

Feeds later layers

  1. L18:312 weakly excites
  2. L19:3022 weakly suppresses
  3. L10:2335 weakly suppresses
  4. L17:1882 weakly suppresses
  5. L9:1755 weakly suppresses
  6. L18:1211 weakly suppresses
  7. L19:1873 weakly suppresses
  8. L11:2635 weakly suppresses
  9. L27:2519 weakly suppresses
  10. L12:983 weakly suppresses unclear: mid-clause words in expository prose
  11. L10:960 weakly excites
  12. L27:398 weakly suppresses
  13. L17:2862 weakly suppresses
  14. L15:2390 weakly excites
  15. L13:378 weakly excites
  16. L10:2475 weakly suppresses
  17. L17:879 weakly excites
  18. L21:2810 weakly suppresses
  19. L10:961 weakly suppresses
  20. L11:321 weakly excites
  21. L10:2144 weakly suppresses
  22. L11:1687 weakly suppresses
  23. L10:1818 weakly suppresses
  24. L22:2871 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 -2.39 and -2.233 values between -1.89 and -1.731 values between -1.73 and -1.561 values between -1.56 and -1.397 values between -1.39 and -1.233 values between -1.23 and -1.067 values between -1.06 and -0.8910 values between -0.89 and -0.7313 values between -0.73 and -0.5646 values between -0.56 and -0.40178 values between -0.40 and -0.23851 values between -0.23 and -0.062273 values between -0.06 and 0.10339 values between 0.10 and 0.2789 values between 0.27 and 0.4440 values between 0.44 and 0.6020 values between 0.60 and 0.7711 values between 0.77 and 0.943 values between 0.94 and 1.104 values between 1.10 and 1.273 values between 1.27 and 1.441 values between 1.77 and 1.941 values between 2.10 and 2.271 values between 2.77 and 2.931 values between 4.10 and 4.27offnegativepositiveoften

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. k
  2. -taking
  3. /no
  4. no
  5. 能找到
  6. ounding
  7. 叶
  8. 谴责
  9. 破裂
  10. clave

Suppresses

  1. cales
  2. сте
  3. antino
  4. raw
  5. adin
  6. manos
  7. テン
  8. contagious
  9. proper
  10. 大事