Neuron Explorer

Qwen3-0.6B

obedience and submission

A neuron in layer 22 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 responds to words and phrases about obeying, following orders or going along with others, such as 'obedience', 'willingly' and 'happily cooperate'. When it was turned up, the model wrote about following rules and obeying, and the effect was clear in several of the test prompts.

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

    that our children should obey our commands, demands and instructions — maintaining a positive and cheery cooperation all the while.⏎Some children do have a compliant temperament and will happily cooperate with requests and commands. This is a successful strategy for getting along in the

  2. #2 peak

    certain rights over their wives in return.⏎Here is a look at some of the rights possessed by a husband in the realm of Islam:⏎1- Obedience and Respect:⏎In the natural scheme of

  3. #3 peak

    control of his own behavior.⏎Authoritative parents discuss and explain their rules and talk about expectations. Even though they do expect their children to follow the rules (not blindly, mind you), they’re also incredibly responsive to their children’s needs. Parent

  4. #4 peak

    7 as examples.)⏎Servants carried out the whims and wishes of the rich. Their lives were dependent primarily on obedience to their employers. Willingly or not, they participated in a continuing economic system that benefitted primarily those who were super-rich.

  5. #5 peak

    parent sets rules and expects the child to follow. That said, the authoritative parent isn’t a dictator. Rules aren’t just randomly given, and children aren’t just told what to do. Instead, parents use age-appropriate boundaries that allow the

  6. #6 peak

    guests. Please divide up this pie between the children.⏎divide and conquer/rule⏎a way of keeping yourself in a position of power by making the people under you disagree with each other so that they are unable to join together and remove you from

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: it is not a problem with the world, but with the way we can make it. This is a very broad and broad clinical and clinical and clinical and / or / / / / / / / / / / / / / / /

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

    onius lies, manipulates people and eaves<|endoftext|>Cancellous bone, also called trabecular bone or spongy bone, light, porous bone enclosing numerous large spaces

  2. #2 peak

    breakfast: Quick, flexible options to grab at home⏎- see all in Healthy menus and shopping strategies⏎Nutritional supplements (3)⏎- Herbal supplements: What to know before you buy⏎- Calcium and calcium supplements: Achieving the right balance

  3. #3 peak

    cater to providing homework help. These websites range from basic free options to paid services. Depending on the site, the quality level of the assistance may vary. Students should do their research and make sure that they are using a site that offers professional help

  4. #4 peak

    00 Web Pages⏎Sample Pages for Prospective Subscribers, or click below⏎Overview of Site|⏎Enchanted Learning Home⏎Monthly Activity Calendar⏎Books to Print⏎Parts of Speech⏎The Test of Time⏎TapQuiz Maps - free

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. L21:2908 suppresses
  2. L20:2171 excites
  3. L21:2416 excites
  4. L21:1163 suppresses 'to' after 'reason' or 'how and when'
  5. L12:1923 suppresses
  6. L19:2320 suppresses
  7. L15:2379 suppresses
  8. L17:2387 suppresses
  9. L20:592 excites
  10. L19:2460 suppresses
  11. L17:2959 excites
  12. L19:1861 suppresses
  13. L16:2426 excites
  14. L17:1986 excites
  15. L21:1183 suppresses
  16. L15:1089 suppresses
  17. L17:3005 weakly excites
  18. L20:2536 weakly excites words for friend or colleague
  19. L17:1254 weakly suppresses
  20. L15:1360 weakly excites
  21. L19:1832 weakly excites
  22. L7:406 weakly excites
  23. L12:3041 weakly excites
  24. L17:2972 weakly suppresses

Feeds later layers

  1. L25:909 weakly excites
  2. L23:2346 weakly excites
  3. L26:966 weakly excites
  4. L25:1388 weakly excites
  5. L23:1199 weakly excites
  6. L25:889 weakly suppresses
  7. L27:1001 weakly excites
  8. L26:2794 weakly excites
  9. L25:1794 weakly excites
  10. L23:1848 weakly excites
  11. L27:2702 weakly excites word fragments before the word ends
  12. L25:370 weakly suppresses
  13. L27:2831 weakly excites
  14. L26:1725 weakly excites
  15. L23:1606 weakly suppresses
  16. L23:615 weakly suppresses
  17. L24:1917 weakly suppresses
  18. L26:1289 weakly excites
  19. L23:661 weakly excites
  20. L25:837 weakly suppresses
  21. L24:1625 weakly excites
  22. L25:1729 weakly excites
  23. L23:199 weakly suppresses
  24. L25:1881 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 -11.56 and -10.761 values between -8.35 and -7.552 values between -5.13 and -4.332 values between -4.33 and -3.5311 values between -3.53 and -2.7212 values between -2.72 and -1.9233 values between -1.92 and -1.12677 values between -1.12 and -0.312766 values between -0.31 and 0.49318 values between 0.49 and 1.2934 values between 1.29 and 2.1013 values between 2.10 and 2.909 values between 2.90 and 3.708 values between 3.70 and 4.519 values between 4.51 and 5.315 values between 5.31 and 6.121 values between 6.12 and 6.922 values between 6.92 and 7.722 values between 7.72 and 8.531 values between 8.53 and 9.331 values between 19.78 and 20.58offnegativepositiveoften

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. 服从
  2. obe
  3. obed
  4. obedience
  5. 听
  6. obey
  7. 遵
  8. 顺
  9. obedient
  10. 聽

Suppresses

  1. _Handler
  2. .raises
  3. osp
  4. 提示
  5. 秾
  6. érique
  7. /Area
  8. strup
  9. thers
  10. št