Neuron Explorer

Qwen3-0.6B

unclear: list items and sentence breaks

A neuron in layer 26 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 is hard to pin down. It fires on a mix of things: list dashes, the end of instruction sentences, and some technical headings. Turning it up made the model output 'En' and 'El' fragments, which does not clearly connect to those examples.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    Rake Angle⏎- Positive and Negative Rake Angle in End Milling⏎- Relationship Between Cutting Tool Angle, Rake Angle, and Relief Angle⏎- Problems in Conventional Rake Angle Measurement⏎- Solution to Problems in Rake Angle

  2. #2 peak

    water, according to the instructions on the label⏎- Use a sponge or rag to apply the detergent solution to the mouldy surface. Scrub the area⏎- Thoroughly dry the surface by<|endoftext|>Home > Technical Articles⏎A person new

  3. #3 peak

    1970 UNESCO Convention on the Means of Prohibiting and Preventing the Illicit Import, Export and Transfer of Ownership of Cultural Property was ratified by the

  4. #4 peak

    according to the U.S. Drug Enforcement Administration (DEA).⏎Ketamine Found Effective, but More Study Suggested⏎The researchers found that ketamine significantly reduced pain scores in 3

  5. #5 peak

    , bladder infections, pneumonia, gonorrhea, and E. coli or salmonella infection.⏎How to use⏎Use Ampicillin as directed by your doctor.⏎Take Ampicillin by mouth at least one-half hour before or 2 hours

  6. #6 peak

    Cover seed to a depth of 1/16 to 1/8 inch by raking it in and lightly⏎rolling or firming the soil.⏎- The finished seedbed should have shallow uniform depressions about⏎1/2 inch

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: what if we reposition the research as a form of research for research? research research research research research research research research research research research research research research research research research research research research research research research research research research research research research research research research research

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

    there is no proof for this. https://www.academia.edu/39903340/Displaying_an_Icon_The_Mosaic_Icon_of_Saint_Demetrios_at_Sassoferrato_and_

  2. #2 peak

    New England Emigrant Aid Company⏎Date: June 9, 1857⏎Charles H. Branscomb, acting on behalf of the New England Emigrant Aid Company, contracted with Abner Doane to construct a school house

  3. #3 peak

    Usage dialog remains open), or|⏎|·||Click on the details for a diagram and click on the Open button to open that diagram and highlight the element (the Element Usage dialog closes)|⏎The material ("material" includes all pages, documents

  4. #4 peak

    (t)=Vmax×sin(ωt+ΦV) and current I(t)=Imax×sin(ωt+ΦI) where Vmax, Imax are the peak values, ω is the common angular pulsation and

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. L19:484 strongly excites
  2. L23:278 strongly excites
  3. L19:474 strongly excites
  4. L25:198 strongly suppresses
  5. L21:185 strongly suppresses
  6. L19:1941 strongly excites
  7. L19:1485 strongly suppresses
  8. L20:351 strongly excites
  9. L20:473 strongly suppresses
  10. L19:2730 strongly suppresses
  11. L21:335 strongly suppresses
  12. L19:1455 strongly suppresses
  13. L21:846 strongly excites
  14. L18:2824 suppresses
  15. L19:96 excites
  16. L25:594 excites
  17. L20:308 excites
  18. L21:370 excites
  19. L24:512 suppresses 'the' inside organisation and convention names
  20. L24:192 suppresses
  21. L19:68 excites full stops closing a claim
  22. L20:1002 suppresses
  23. L20:296 suppresses
  24. L22:1829 excites

Feeds later layers

  1. L27:2632 suppresses
  2. L27:387 suppresses 'a' right after 'played' or 'play'
  3. L27:455 excites
  4. L27:53 suppresses
  5. L27:1745 excites
  6. L27:1507 excites
  7. L27:137 weakly excites
  8. L27:715 weakly excites
  9. L27:2097 weakly excites
  10. L27:787 weakly excites
  11. L27:16 weakly excites
  12. L27:49 weakly suppresses
  13. L27:601 weakly excites
  14. L27:1599 weakly excites
  15. L27:12 weakly excites
  16. L27:918 weakly excites
  17. L27:229 weakly excites
  18. L27:1230 weakly excites
  19. L27:852 weakly excites
  20. L27:133 weakly excites
  21. L27:822 weakly excites
  22. L27:316 weakly excites
  23. L27:178 weakly suppresses
  24. L27:1361 weakly excites

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 -19.03 and -18.061 values between -18.06 and -17.101 values between -16.13 and -15.161 values between -15.16 and -14.191 values between -13.22 and -12.261 values between -11.29 and -10.322 values between -10.32 and -9.353 values between -9.35 and -8.385 values between -8.38 and -7.427 values between -7.42 and -6.457 values between -6.45 and -5.4810 values between -5.48 and -4.5116 values between -4.51 and -3.5427 values between -3.54 and -2.5858 values between -2.58 and -1.61187 values between -1.61 and -0.642882 values between -0.64 and 0.33361 values between 0.33 and 1.30121 values between 1.30 and 2.2661 values between 2.26 and 3.2347 values between 3.23 and 4.2038 values between 4.20 and 5.1717 values between 5.17 and 6.1413 values between 6.14 and 7.1017 values between 7.10 and 8.077 values between 8.07 and 9.044 values between 9.04 and 10.016 values between 10.01 and 10.982 values between 10.98 and 11.941 values between 11.94 and 12.911 values between 12.91 and 13.881 values between 13.88 and 14.851 values between 18.72 and 19.69offnegativepositiveoften

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. En
  2. El
  3. Es
  4. El
  5. O
  6. Ex
  7. Em
  8. O
  9. Ed
  10. Input

Suppresses

  1. atory
  2. ämp
  3. asting
  4. aturing
  5. utory
  6. ucky
  7. ทน
  8. fähig
  9. سبة
  10. ないので