Neuron Explorer

Qwen3-0.6B

'have' after 'you' or 'to'

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 responds to the word 'have' when it comes right after 'you' or 'to', as in 'make sure you have' or 'once you have the value'. It is strongly suppressed by 'take on' in phrases like 'never takes on the same value'. Turning it up made the text wander into unrelated, rambling sentences.

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

    to find the value of the mark. You can do this by using a percentage calculator. You can use it to find the value of a number. Once you have the value, you can write it down in terms of the percent. It’s

  2. #2 peak

    service of: ASD at NASA / GSFC⏎& Michigan Tech. U.<|endoftext|>Troubleshooting With Superheat, Subcooling⏎You also need to have the right tools and know-how to apply these principles to use the tool quickly and

  3. #3 peak

    – dd will copy every bit of it, even if there’s only one tiny file stored on the whole hard drive. So, to start, make sure you have a device connected to your computer that will be

  4. #4 peak

    one time, make sure you have all the supplies you need at hand.⏎A lot of nurseries, garden centres

  5. #5 peak

    . Test the alarm often to be sure it’s working properly. Replace the entire smoke detector every 10 years.⏎- Make sure you have enough smoke detectors in your house. Position them inside or just outside bedrooms, in

  6. #6 peak

    are the part of same Microsoft exchange server.⏎How do I know if a recalled communication was successful in outlook?⏎Make sure you have to tick the following option when recalling the email. Outlook will send you a notice

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 position of a single person in a position of the form of a single person in the form of a single person in the form of a single person in the form of a single person in the form of a single person in the form of

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

    with are called one-to-one functions. DEFINITION A function is called a one-to-one function if it never takes on the same value twice; that is, If a horizontal line intersects the graph of

  2. #2 peak

    1 FUNCTIONS AND MODELS Not all functions possess inverses. Lets compare the functions and whose arrow diagrams are shown in Figure 1. Note that never takes on the same value twice (any two inputs in have different outputs), whereas does take

  3. #3 peak

    Gambling operates according to conditioning principles, with emotional and social rewards playing a role as well as financial ones. The sights, sounds and other stimuli associated with gambling take on some of the same rewarding properties and thereby act as incentives in themselves, deepening

  4. #4 peak

    until this point, his paintings were of peasant life in sombre colours. Once settled in Paris, however, his paintings took on new painting styles and brighter colours, and he started painting urban and suburban subjects.

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. L3:2551 weakly excites
  2. L3:2096 weakly excites
  3. L1:325 weakly excites
  4. L0:963 weakly excites
  5. L0:1026 weakly excites
  6. L2:1095 weakly suppresses
  7. L3:2249 weakly excites
  8. L1:2335 weakly excites
  9. L3:632 weakly suppresses
  10. L3:1609 weakly excites
  11. L0:1920 weakly suppresses
  12. L3:1057 weakly suppresses
  13. L2:2818 weakly suppresses
  14. L3:1218 weakly suppresses
  15. L3:754 weakly excites
  16. L2:2804 weakly suppresses
  17. L3:59 weakly excites
  18. L0:1447 weakly excites
  19. L0:1347 weakly excites
  20. L2:546 weakly excites
  21. L3:2859 weakly excites
  22. L0:2163 weakly suppresses
  23. L3:878 weakly suppresses 'of' after 'Because'
  24. L3:2140 weakly suppresses

Feeds later layers

  1. L8:2764 suppresses
  2. L12:1475 excites
  3. L9:2196 weakly suppresses
  4. L14:1017 weakly suppresses
  5. L15:2682 weakly suppresses
  6. L5:267 weakly suppresses
  7. L23:1481 weakly suppresses
  8. L6:2310 weakly suppresses
  9. L15:1223 weakly suppresses
  10. L15:1951 weakly excites
  11. L21:2626 weakly excites
  12. L10:886 weakly suppresses
  13. L14:1219 weakly suppresses
  14. L7:2420 weakly suppresses
  15. L17:1107 weakly suppresses
  16. L8:2458 weakly excites
  17. L9:1795 weakly suppresses
  18. L9:2164 weakly suppresses
  19. L9:1748 weakly suppresses
  20. L5:2354 weakly excites
  21. L14:2776 weakly excites
  22. L7:2541 weakly suppresses
  23. L5:2898 weakly suppresses
  24. L16:2286 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 -3.08 and -2.971 values between -2.29 and -2.181 values between -1.73 and -1.621 values between -1.62 and -1.513 values between -1.51 and -1.401 values between -1.40 and -1.283 values between -1.28 and -1.171 values between -1.06 and -0.953 values between -0.95 and -0.833 values between -0.83 and -0.723 values between -0.72 and -0.617 values between -0.61 and -0.509 values between -0.50 and -0.3948 values between -0.39 and -0.27204 values between -0.27 and -0.16609 values between -0.16 and -0.051409 values between -0.05 and 0.061109 values between 0.06 and 0.18368 values between 0.18 and 0.2997 values between 0.29 and 0.4012 values between 0.40 and 0.518 values between 0.51 and 0.633 values between 0.63 and 0.741 values between 0.96 and 1.071 values between 1.19 and 1.302 values between 1.30 and 1.41offnegativepositiveoften

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. aktion
  2. filtr
  3. proj
  4. 滤
  5. 捣
  6. jal
  7. 炉
  8. 담
  9. 终
  10. 在哪里

Suppresses

  1. زا
  2. ITIVE
  3. "He
  4. ductive
  5. казал
  6. ests
  7. “He
  8. hữu
  9. natural
  10. itated