Neuron Explorer

Qwen3-0.6B

unclear: scattered technical nouns

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 weakly on ordinary nouns in technical or science text, such as 'network', 'court', 'cannabis' and 'cortisol'. There is no obvious single theme. Turning it up gave writing with no clear link to these words.

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

    limited applicability in long-term tests. Tracers may be roughly classified as passive or active. In principle, a passive tracer blindly follows the fluid phase in which it is injected. Interpretation of tracer-production curves must account for this. The results

  2. #2 peak

    ⏎CS610: Computer Networks⏎A network is defined as a system for connecting computers using a single transmission technology. Computers can communicate with others on the network. They can send and receive data from each other when they are in

  3. #3 peak

    or so employees. The district courts have a local link - the cases that are considered by the district court come from those municipalities that fall within the district of the court (a geographical area that is included in the district court´s catchment area,

  4. #4 peak

    effective at freezing stuff than water ice at 0 C. You can make ice cream, just as with liquid nitrogen, by adding dry ice directly to the ice cream mixture. However, because dry ice is frozen carbon dioxide, this procedure results in

  5. #5 peak

    55. Let the maximum value for pixels in image I be MAX.⏎Express the PSNR in decibels. From Step 1, we have the decibel value LdB as LdB = 10 log10 (P

  6. #6 peak

    things, for consumption and to make users ‘high’. Thus, we can safely call marijuana a drug.⏎To start with, marijuana, which can come from the cannabis Indica or Sativa species, is the dry flower of the female hemp plant

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 aware of the fact that I am a user of the system. So, I have to make sure that my user account is safe and that I can use it. This is my first step, which is to create a

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

    illes’ home.⏎- The doctor says that Louisa will recover, but cannot be moved.⏎- Louisa slowly recovers in Lyme.⏎- As she recovers, Louisa develops a taste for poetry and an affection for Captain Benwick.⏎

  2. #2 peak

    and Santa Rita Mountain ranges outside Patagonia, AZ.⏎Obstacles to the Protection of Wildlife Corridors⏎Given that the most important linkages have been identified, why has so little been done to secure wildlife corridors in this region? There

  3. #3 peak

    government is headed by the President, who appoints a vice-president and ministers as head of state. The President also exercises some powers that are vested in the National Assembly. In case there is no clear majority on important issues facing this country, it

  4. #4 peak

    5.2 or under drought conditions. The tubers are most susceptible, but stems and roots may occasionally be infected. No above-ground symptoms are apparent on infected potato plants and scab does not adversely affect potato tubers in storage.⏎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. L7:1185 suppresses
  2. L7:934 excites
  3. L6:484 weakly excites
  4. L7:336 weakly suppresses
  5. L5:144 weakly excites
  6. L7:1023 weakly suppresses
  7. L7:158 weakly suppresses
  8. L7:618 weakly suppresses
  9. L3:147 weakly excites
  10. L7:215 weakly excites
  11. L5:2303 weakly excites
  12. L7:1183 weakly excites
  13. L7:902 weakly excites
  14. L7:667 weakly excites
  15. L7:585 weakly excites
  16. L7:741 weakly excites
  17. L4:24 weakly suppresses
  18. L6:198 weakly excites
  19. L1:317 weakly suppresses
  20. L6:568 weakly suppresses
  21. L6:769 weakly excites
  22. L1:2220 weakly suppresses
  23. L7:2302 weakly excites
  24. L6:409 weakly excites

Feeds later layers

  1. L11:592 excites
  2. L13:1668 suppresses
  3. L9:313 excites
  4. L12:517 excites
  5. L10:1412 excites
  6. L11:741 suppresses
  7. L10:1763 suppresses
  8. L11:2447 excites
  9. L9:1073 suppresses
  10. L11:1225 suppresses
  11. L12:300 excites
  12. L13:181 suppresses
  13. L11:1126 suppresses
  14. L9:1096 excites
  15. L14:256 excites
  16. L9:898 excites
  17. L15:2972 suppresses
  18. L11:1070 suppresses
  19. L11:1756 suppresses
  20. L12:1136 suppresses
  21. L10:867 excites
  22. L10:368 suppresses
  23. L9:509 suppresses
  24. L11:3069 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 -1.95 and -1.881 values between -1.42 and -1.341 values between -0.96 and -0.881 values between -0.88 and -0.812 values between -0.81 and -0.734 values between -0.73 and -0.651 values between -0.58 and -0.507 values between -0.50 and -0.4316 values between -0.43 and -0.3543 values between -0.35 and -0.27111 values between -0.27 and -0.20291 values between -0.20 and -0.12570 values between -0.12 and -0.041302 values between -0.04 and 0.03802 values between 0.03 and 0.11428 values between 0.11 and 0.19178 values between 0.19 and 0.2692 values between 0.26 and 0.3440 values between 0.34 and 0.4111 values between 0.41 and 0.493 values between 0.49 and 0.571 values between 0.57 and 0.641 values between 0.64 and 0.721 values between 1.02 and 1.10offnegativepositiveoften

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. .Ptr
  2. 什么是
  3. ourke
  4. ibur
  5. 的决心
  6. 新增
  7. 篷
  8. yb
  9. احة
  10. 的心理

Suppresses

  1. 庫
  2. 安然
  3. 칩
  4. amines
  5. ubes
  6. quisite
  7. ucchini
  8. OKEN
  9. uite
  10. ி