Neuron Explorer

Qwen3-0.6B

'addition' in 'In addition to'

A neuron in layer 19 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 fires on the word 'addition' when it begins a sentence as 'In addition', right before the model would write 'to' or a comma. Its negative side is a different, unrelated pattern: a plus sign inside time zones and phone numbers. Steering did not produce a clear related theme.

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

    ylinders⏎Owing to the above discussed hazards, it has become important to take special care while storing and handling compressed gas cylinders of acetylene.⏎In addition to the above-mentioned

  2. #2 peak

    with the personal details like age, address and choices in products and services make a unique combination<|endoftext|>Today we take a closer look at the history of yogurt. In addition to being a delicious source of nutrients, this dairy product is also a mine of

  3. #3 peak

    suggest the rest. Bisbee's intuitive working method comes closest to the surrealist process of automatic drawing, allowing the subconscious mind to direct the hand.⏎In addition to the "Tons" series, Bisbee has recently produced more intricately

  4. #4 peak

    keys in hotels. This type of technology can help employees to check in and out of rooms more efficiently, which could mean reduced staffing costs and reduced interaction.⏎In addition to be a cost-effective way to save money, IoT technology can also be used

  5. #5 peak

    , facial numbness, etc. Conversely, if the tumor is found affecting the vestibular nerves of both ears, you can expect symptoms on both sides.⏎In addition to the symptoms of an acoustic neuroma, if the tumor grows large enough to

  6. #6 peak

    development of nozzle technology, the water spray posture has changed a lot. With these materials, through professional art design, you can draw a beautiful aquatic landscape.⏎In addition, different creative fountain forms are suitable for different application scenarios. For example, music

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 the best possible, and I need to be the best possible in every aspect of my life. So I feel that I have to be a person who is capable of being the best possible. I feel that I am 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

    olving in the lab’?⏎‘A poke in the eye for anti-evolutionists’?⏎Published: 14 June 2008 (GMT+10)⏎Some bacteria cultured in a laboratory have gained the ability to use cit

  2. #2 peak

    -(0)48-503-9576 / Fax: +81-(0)48-503-9609

  3. #3 peak

    Seafood producers tackle ocean acidification at annual Summit⏎Monday, February 21, 2011, 23:50 (GMT + 9)⏎From 31 January to 2 February, more than

  4. #4 peak

    strip supply by 2030 as world population grows and demand for power rises, according to the 2030 water resources group:⏎Phone: +32 2 213 18 33⏎E

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. L17:341 suppresses
  2. L0:2580 suppresses
  3. L16:87 suppresses
  4. L3:348 suppresses
  5. L18:253 suppresses
  6. L16:463 excites 'for' introducing a purpose or target
  7. L5:163 suppresses
  8. L4:303 suppresses
  9. L9:327 excites
  10. L18:569 excites
  11. L16:76 suppresses
  12. L15:96 suppresses
  13. L15:114 excites
  14. L4:2771 suppresses
  15. L16:282 excites
  16. L4:20 weakly excites
  17. L4:1360 weakly suppresses
  18. L15:78 weakly excites
  19. L15:89 weakly excites
  20. L15:80 weakly suppresses
  21. L6:726 weakly suppresses
  22. L16:1558 weakly suppresses
  23. L16:493 weakly excites
  24. L15:129 weakly suppresses

Feeds later layers

  1. L20:2375 strongly suppresses
  2. L27:1007 suppresses
  3. L23:1479 suppresses
  4. L21:2896 suppresses
  5. L21:463 excites
  6. L20:197 weakly excites
  7. L20:180 weakly excites
  8. L25:679 weakly excites
  9. L21:280 weakly excites
  10. L27:692 weakly excites 'Ch' starting a proper name
  11. L27:1332 weakly excites
  12. L26:1489 weakly excites
  13. L22:428 weakly excites
  14. L20:97 weakly excites
  15. L20:216 weakly excites
  16. L20:221 weakly excites
  17. L27:1769 weakly excites "p" starting a word mid-token
  18. L22:215 weakly excites
  19. L22:1298 weakly suppresses
  20. L23:308 weakly excites
  21. L20:149 weakly excites
  22. L20:417 weakly excites
  23. L22:979 weakly excites
  24. L21:2887 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 -8.46 and -8.201 values between -4.74 and -4.471 values between -4.20 and -3.941 values between -3.94 and -3.671 values between -3.41 and -3.141 values between -3.14 and -2.871 values between -2.87 and -2.611 values between -2.61 and -2.341 values between -2.34 and -2.073 values between -2.07 and -1.811 values between -1.81 and -1.541 values between -1.54 and -1.284 values between -1.28 and -1.017 values between -1.01 and -0.746 values between -0.74 and -0.485 values between -0.48 and -0.21750 values between -0.21 and 0.061673 values between 0.06 and 0.32701 values between 0.32 and 0.59366 values between 0.59 and 0.85192 values between 0.85 and 1.12104 values between 1.12 and 1.3959 values between 1.39 and 1.6522 values between 1.65 and 1.925 values between 1.92 and 2.18offnegativepositiveoften

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. seys
  2. +W
  3. UIL
  4. ��
  5. .US
  6. replaced
  7. legen
  8. anism
  9. +E
  10. headquarters

Suppresses

  1. حل
  2. kidd
  3. slash
  4. ifications
  5. �
  6. gmail
  7. �
  8. flip
  9. osob
  10. endum