Neuron Explorer

Qwen3-0.6B

"in some" opening a qualifier

A neuron in layer 4 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 "some" when it comes right after "In" or "At" to open a sentence like "In some cases" or "In some countries". Turning it up gave wandering text that did not clearly reflect this.

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

    are published in a state’s code are grouped by subject matter into Titles, as in Title 11 of the United States Code (bankruptcy laws). In some states, including California, Texas and New York, the term "code" may

  2. #2 peak

    have often been without and are looking for a provider to assure their future financially. We have also learned to patiently build trust instead of expecting it. At times even enduring undeserved blame for not being the expected hero, meeting their needs as

  3. #3 peak

    protection[change | change source]⏎Copyright laws usually protect owners of copyright beyond their lifetime, such as until 50 years after the author's death. In some countries this period has been extended to 70 years. When the period of

  4. #4 peak

    involve an index of software patches and upgrades for key parts of the software system. Experts also recommend vulnerability scans that look for risk factors across complex software systems. In some cases, this involves a multi-phase process that consists of multiple scans and remediation

  5. #5 peak

    2 to 18 inches of soil, although the depth depends on a site's soil conditions. Root systems usually are shallower in heavy clay soils.⏎In some situations, it can be detrimental to amend or replace heavy clay backfill soil,

  6. #6 peak

    7799 for IT security and the ISO 9000 series for management, establish guidelines and general principles that codify industry best practices. In some cases independent certification

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 need to be able to find the input to the first line of the given here. So, 0 1 2 3 4 5 6 Okay, I need to be able to find the input for the

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

    Bagh, Sheesh Mahal, Saif-ul-Maluk Lake, Mohenjo-daro, Minar-e-Pakistan, Pakistan Monument.⏎The main city

  2. #2 peak

    Indus Valley Civilisation of Harappa and Mohenjo-daro from the Bronze Age. Nowadays, the Thar Desert is the

  3. #3 peak

    Atmospheric Release Advisory Center in Livermore also assisted with the 1986 Chernobyl nuclear power plant disaster and the 1979 Three Mile Island accident

  4. #4 peak

    , sulfur, and microorganisms. (Crude oil with lots of sulfur in it is called “sour,” as opposed to “sweet.”) If enough of these contaminants collect in a steel pipe, they’ll work together to eat

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. L3:122 excites
  2. L1:1340 weakly suppresses
  3. L1:1029 weakly excites
  4. L3:464 weakly excites
  5. L3:186 weakly suppresses
  6. L3:1945 weakly excites 'to' in 'person to person' phrases
  7. L2:1176 weakly suppresses
  8. L3:1394 weakly excites
  9. L3:512 weakly suppresses
  10. L3:914 weakly excites "any" in negated sentences
  11. L2:549 weakly suppresses
  12. L1:2801 weakly excites
  13. L3:50 weakly excites
  14. L2:1127 weakly suppresses
  15. L0:1870 weakly excites
  16. L3:342 weakly excites
  17. L1:279 weakly excites
  18. L3:3060 weakly suppresses
  19. L3:691 weakly excites
  20. L2:1315 weakly excites 'of' after nouns like 'variation' and 'section'
  21. L2:2328 weakly suppresses
  22. L2:2834 weakly excites
  23. L3:2181 weakly suppresses comma inside numbers
  24. L2:1124 weakly excites

Feeds later layers

  1. L11:522 excites
  2. L7:2733 excites
  3. L8:2231 excites
  4. L15:2651 suppresses
  5. L18:2896 excites
  6. L16:2830 excites
  7. L7:561 suppresses 'all' or 'every' before a group
  8. L21:331 excites
  9. L13:2339 excites
  10. L15:869 excites
  11. L21:1472 weakly excites
  12. L19:2909 weakly suppresses
  13. L14:1763 weakly excites
  14. L5:85 weakly suppresses
  15. L16:311 weakly suppresses
  16. L8:981 weakly excites
  17. L9:352 weakly excites
  18. L14:1175 weakly excites
  19. L11:158 weakly suppresses
  20. L12:942 weakly suppresses
  21. L17:847 weakly excites
  22. L6:220 weakly excites
  23. L19:2973 weakly suppresses
  24. L10:2867 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 -1.76 and -1.421 values between -1.08 and -0.7442 values between -0.74 and -0.401792 values between -0.40 and -0.052017 values between -0.05 and 0.2932 values between 0.29 and 0.634 values between 0.63 and 0.974 values between 0.97 and 1.312 values between 1.65 and 1.991 values between 2.33 and 2.681 values between 3.36 and 3.701 values between 3.70 and 4.041 values between 4.04 and 4.381 values between 4.38 and 4.721 values between 6.43 and 6.771 values between 8.47 and 8.812 values between 9.15 and 9.501 values between 9.50 and 9.841 values between 9.84 and 10.181 values between 10.52 and 10.861 values between 11.54 and 11.88offnegativepositiveoften

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. leet
  2. mund
  3. 需求
  4. RequestId
  5. greSQL
  6. 挑战
  7. ye
  8. 来电
  9. entimes
  10. 钽

Suppresses

  1. IMO
  2. %+
  3. Główn
  4. 深交
  5. /=
  6. Comparable
  7. endforeach
  8. xious
  9. closest
  10. สะ