Neuron Explorer

Qwen3-0.6B

unclear: commas and 'or' in lists

A neuron in layer 8 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 is weak and hard to pin down. Its strongest hits are commas and words like 'or' inside lists or clauses, but the pattern is not consistent. Turning it up did not produce a clear theme.

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

    contact, while 40% reported experiencing verbal abuse.⏎The problem is not limited to female students, however. Male students also experience sexual harassment, although to a lesser extent. According to the same study, 20% of male university

  2. #2 peak

    foods that contain gluten. Gluten is a protein present in a certain number of cereals: wheat, rye, oats and barley. Foods containing these cereals or their derivatives,

  3. #3 peak

    same output.⏎- Nearly Every Database supports available⏎One of the greatest features in PHP is its support for a wide range of databases, nearly every databases – MySQL, ORACLE, PostgreSQL, Sybase, Informix to name a few.⏎What

  4. #4 peak

    umbled.⏎Walesa called for that persistence from his listeners. The message was only slightly different from his speech four years earlier at a College-sponsored event attended by, among others, local mayors and elected officials.⏎Things have been changing in Cuba

  5. #5 peak

    cted a description of liberty to justify it. The “devil,” in this case, lies not in the details but in the harm principle:⏎That principle is, that the sole end for which mankind are warranted, individually or collectively in interfering with

  6. #6 peak

    local Excessive Heat Warning criteria in the next 24 to 72 hours.⏎- Excessive Heat Warning⏎– Heat Index values are forecasting to meet or exceed locally defined warning criteria for at least 2 days (daytime highs=

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 have a 43-year-old man who is a former partner in the past. The past 20 years have passed, and I have been a member of the organization, and I have been a member of the organization. I

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

    unda tamamlamıştır.⏎Yüksek Lisans Öğrenimine MSÜ Barbaros DEBİM Enstitüsünde devam etmekte aynı zamanda enerji sistem ve modelleme üzerine araştırmalar

  2. #2 peak

    permitted the right to gain literacy’ entrenched so deeply in people’s mindset that it resulted in a lacking of reliable narrative sources.⏎That Douglass was about to write drew public attention, put him under the limelight, and made him the only loud

  3. #3 peak

    Summer in North India is indeed hot and cruel, the long hot and often humid summer days, and the even more uncomfortable sultry, hot evenings in the capital leave us all dehydrated and exhausted. The heat, and the air conditioning,

  4. #4 peak

    the jet age and the information age. Geography is where any discussion of the strategic environment must begin. To know a country’s geography provides clues to its intentions better than any wiretap of its national security meetings. The geography that concerned us in the

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. L7:554 suppresses
  2. L6:335 suppresses
  3. L7:152 suppresses
  4. L4:153 suppresses
  5. L6:65 suppresses
  6. L5:433 suppresses
  7. L4:43 suppresses
  8. L6:145 excites
  9. L7:1582 suppresses
  10. L4:2955 excites
  11. L7:1584 suppresses
  12. L7:3018 suppresses
  13. L6:230 weakly suppresses
  14. L6:541 weakly suppresses
  15. L7:506 weakly excites
  16. L6:1648 weakly excites
  17. L6:393 weakly suppresses
  18. L4:735 weakly suppresses
  19. L7:223 weakly suppresses
  20. L7:69 weakly suppresses
  21. L3:853 weakly suppresses
  22. L4:91 weakly excites
  23. L3:73 weakly suppresses
  24. L7:362 weakly excites

Feeds later layers

  1. L10:319 excites
  2. L9:544 excites
  3. L12:1222 excites
  4. L9:1196 excites
  5. L9:680 excites
  6. L15:1746 excites
  7. L9:1143 excites
  8. L13:2988 excites
  9. L10:1781 excites
  10. L9:738 suppresses
  11. L11:2655 excites
  12. L11:775 excites
  13. L14:2088 excites
  14. L14:1920 excites
  15. L9:686 excites
  16. L11:1290 excites
  17. L9:1088 excites
  18. L13:76 suppresses
  19. L12:2513 excites
  20. L14:1227 excites
  21. L13:1385 excites
  22. L10:899 excites
  23. L10:756 excites
  24. L14:2204 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 -2.51 and -2.401 values between -2.09 and -1.991 values between -1.99 and -1.881 values between -1.88 and -1.784 values between -1.78 and -1.671 values between -1.67 and -1.571 values between -1.57 and -1.471 values between -1.47 and -1.368 values between -1.36 and -1.265 values between -1.26 and -1.154 values between -1.15 and -1.059 values between -1.05 and -0.9413 values between -0.94 and -0.8420 values between -0.84 and -0.7426 values between -0.74 and -0.6344 values between -0.63 and -0.5352 values between -0.53 and -0.4276 values between -0.42 and -0.32124 values between -0.32 and -0.21238 values between -0.21 and -0.11637 values between -0.11 and -0.01945 values between -0.01 and 0.10491 values between 0.10 and 0.20361 values between 0.20 and 0.31266 values between 0.31 and 0.41180 values between 0.41 and 0.52124 values between 0.52 and 0.6282 values between 0.62 and 0.7375 values between 0.73 and 0.8334 values between 0.83 and 0.9336 values between 0.93 and 1.0418 values between 1.04 and 1.1410 values between 1.14 and 1.258 values between 1.25 and 1.352 values between 1.35 and 1.465 values between 1.46 and 1.564 values between 1.56 and 1.66offnegativepositiveoften

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. otive
  2. Subtitle
  3. &,
  4. hs
  5. Solar
  6. cular
  7. lsi
  8. DUCTION
  9. efs
  10. hält

Suppresses

  1. ochen
  2. ivor
  3. _possible
  4. eren
  5. 权
  6. 所所长
  7. 所
  8. .prototype
  9. inator
  10. fce