Neuron Explorer

Qwen3-0.6B

'up to' meaning someone's choice

A neuron in layer 7 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 responds to the phrase 'up to', as in 'it is up to you' or 'up to the players to decide'. It is about leaving a decision to someone else. Turned up, the model's text moved toward choices, personal responsibility and deciding for yourself.

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

    GameMaster (also written DM and GM), directs one or more players through a made-up world. Typically, the GM will describe a situation and it is up to the players to decide how their characters will react.⏎An extension to this is a

  2. #2 peak

    throughout history of America.” Leaving the specifics up to a teacher or LEA who have only been educated in elementary education is ludicrous.

  3. #3 peak

    river and its conditions.⏎Boating alone is unsafe. Two or more people are recommended.⏎Know first aid.⏎Be prepared for self-rescue. It is up to you and everyone in your boat to be able to help.⏎View River Runner Et

  4. #4 peak

    it will vary from home to home, depending on your family's values. School rules, don't encompass every single behavior. In most instances it will be up to your children's personal<|endoftext|>Science & Medicine⏎About the Photo⏎Outdoor yoga class

  5. #5 peak

    or they haven’t had the proper training. It’s up to the manager from this point to ensure the employee’s contributions are valued and that their

  6. #6 peak

    of 1979, “If we cannot celebrate a man who died for love, then how can we say we believe in it? It is up to me and you.”⏎Adafruit publishes a wide range of writing and video content,

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: the number 10, 1000, 10000000000000000000000000000000000

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

    a child’s immune system is functioning. Specifically, the test measures the levels of certain autoantibodies (responsible for causing the symptoms) circulating in a child’s bloodstream

  2. #2 peak

    arranged in a deep architecture to model the high level abstraction present in the data . There are many deep learning techniques currently used in various applications. These include auto-encoders, stacked auto-encoders, restricted Boltzmann machines (RBM

  3. #3 peak

    hormonal) control is called ............... . This enables the nephrons to maintain a stable GFR even when arterial blood pressure changes.⏎Which mechanism of renal autoregulation responds to an INCREASE in BP which causes an increased STRETCH in

  4. #4 peak

    function. The surgical options include:⏎- Marrow stimulation to help encourage the growth of new cartilage⏎- Debridement to remove loose edges⏎- Autologous chondrocyte implantation to create new cartilage and replace what

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. L6:1977 suppresses
  2. L6:1118 suppresses
  3. L5:2633 excites
  4. L6:921 excites
  5. L6:2614 weakly suppresses
  6. L5:1091 weakly suppresses
  7. L4:2101 weakly suppresses
  8. L6:2934 weakly suppresses
  9. L4:1504 weakly excites
  10. L4:2707 weakly suppresses
  11. L5:2480 weakly excites
  12. L4:2779 weakly excites
  13. L2:2638 weakly suppresses
  14. L3:2350 weakly suppresses
  15. L1:1751 weakly excites
  16. L6:2262 weakly suppresses
  17. L6:1949 weakly excites
  18. L6:2914 weakly suppresses
  19. L6:2284 weakly suppresses
  20. L5:1940 weakly suppresses
  21. L4:1407 weakly suppresses
  22. L5:845 weakly suppresses
  23. L6:1190 weakly excites
  24. L0:1637 weakly suppresses

Feeds later layers

  1. L10:223 excites
  2. L12:2806 excites
  3. L20:2946 excites
  4. L15:1644 excites
  5. L9:3051 excites
  6. L8:2656 weakly excites
  7. L12:2388 weakly excites
  8. L12:2527 weakly excites
  9. L14:1123 weakly excites
  10. L11:2186 weakly excites
  11. L11:1959 weakly excites
  12. L13:2056 weakly excites
  13. L14:1480 weakly excites
  14. L14:1684 weakly excites
  15. L15:1399 weakly excites
  16. L14:2687 weakly excites
  17. L11:1205 weakly excites
  18. L12:2511 weakly excites
  19. L13:2649 weakly excites
  20. L11:636 weakly suppresses
  21. L12:2482 weakly excites
  22. L9:1519 weakly excites
  23. L10:2675 weakly excites
  24. L27:3024 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.

2 values between -1.52 and -1.381 values between -1.10 and -0.961 values between -0.82 and -0.685 values between -0.68 and -0.5435 values between -0.54 and -0.40154 values between -0.40 and -0.26689 values between -0.26 and -0.121581 values between -0.12 and 0.021025 values between 0.02 and 0.16323 values between 0.16 and 0.3062 values between 0.30 and 0.4415 values between 0.44 and 0.585 values between 0.58 and 0.724 values between 0.72 and 0.861 values between 1.14 and 1.282 values between 1.56 and 1.701 values between 1.98 and 2.121 values between 2.12 and 2.261 values between 3.93 and 4.07offnegativepositiveoften

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. 意志
  2. 主观
  3. subjective
  4. Choice
  5. 选择
  6. independently
  7. ensual
  8. choices
  9. dio
  10. choice

Suppresses

  1. ارية
  2. 打进
  3. rá
  4. burge
  5. &$
  6. ví
  7. ernal
  8. 千亿
  9. %@",
  10. otyping