Neuron Explorer

Qwen3-0.6B

'with' in 'along with'

A neuron in layer 1 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 'with' in the phrase 'along with', which adds an extra item to a list. The negative side is weak and shows up on the 'JC' part of journal subject codes. Turning it up did not make the writing obviously about this phrase.

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

    commercial marketing, and large quantities of unused opioids made easily accessible in the home” (National Academy of Medicine, 2017)⏎These factors, along with the scientific literature perspective that opioids produced low rates of addiction and were relatively safe,

  2. #2 peak

    These kinds of statistics, along with their experience of using the Pinard horn, got the three computer science students thinking

  3. #3 peak

    destined to be drastically lower. This, along with its direct effect on a reduction in energy consumption, makes the new technology a potential

  4. #4 peak

    practices and ensure the protection of conservation targets.⏎The Cayos Cochinos Foundation, along with the Conservancy and other groups, has also worked with the communities to establish management

  5. #5 peak

    high-resolution photographs of each weevil that will be uploaded to Species ID, along with a short scientific description. More than 100 species were brought to the

  6. #6 peak

    Congress, along with the debates over the adoption of the Bill of Rights.⏎Copied below is the September 1787 publication of the Constitution, along with Federalist I , which appeared in

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 have been thinking about the relationship between the two things, and I have been thinking about the relationship between the two things. Let me think of the relationship between the two things. I have been thinking about the relationship between the two things,

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

    results demonstrate that silkworm-BEVS is a promising alternative strategy to produce the functional and thermostable reverse transcriptase.⏎All Science Journal Classification (ASJC) codes⏎- Insect Science<|endoftext|>An account of the religion of

  2. #2 peak

    -layer coated, triple-layer coated and graded-index-layer coated samples was performed and compared with experimentally derived films. Good agreement between model and experiment was observed.⏎ASJC Scopus subject areas⏎- Electronic, Optical and Magnetic Materials⏎- Ceramics

  3. #3 peak

    |Number of pages||12|⏎|Journal||Classical and Quantum Gravity|⏎|State||Published - Jul 7 2003|⏎ASJC Scopus subject areas⏎- Physics and Astronomy (miscellaneous)<|endoftext|>Most computer

  4. #4 peak

    diverse set of controls, as well as in a discontinuity analysis on observations near the borders between anglophone and francophone countries.⏎- Colonial rule⏎ASJC Scopus subject areas⏎- Economics and Econometrics⏎- Organizational Behavior and

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. L0:178 suppresses
  2. L0:1224 suppresses
  3. L0:1301 suppresses
  4. L0:1 suppresses
  5. L0:563 weakly suppresses 'www' in web addresses
  6. L0:799 weakly suppresses
  7. L0:2666 weakly suppresses
  8. L0:659 weakly suppresses
  9. L0:2427 weakly suppresses
  10. L0:2021 weakly excites
  11. L0:17 weakly suppresses
  12. L0:2682 weakly excites
  13. L0:2377 weakly excites
  14. L0:161 weakly excites
  15. L0:1985 weakly excites
  16. L0:1433 weakly suppresses
  17. L0:1158 weakly excites
  18. L0:32 weakly suppresses
  19. L0:768 weakly suppresses
  20. L0:1094 weakly suppresses
  21. L0:2487 weakly excites
  22. L0:2812 weakly excites
  23. L0:423 weakly excites
  24. L0:346 weakly suppresses comma closing an introductory phrase

Feeds later layers

  1. L3:250 excites
  2. L24:1006 excites
  3. L3:538 weakly excites
  4. L8:2614 weakly excites
  5. L20:2343 weakly excites
  6. L19:1236 weakly excites
  7. L26:2171 weakly excites
  8. L6:792 weakly suppresses
  9. L5:1469 weakly excites
  10. L13:359 weakly suppresses
  11. L8:1083 weakly excites "x" or "by" between two dimensions
  12. L23:1976 weakly suppresses
  13. L19:177 weakly suppresses
  14. L18:2320 weakly excites
  15. L22:1877 weakly excites
  16. L3:704 weakly excites
  17. L13:563 weakly suppresses
  18. L23:2181 weakly excites
  19. L7:3071 weakly suppresses
  20. L11:251 weakly excites
  21. L7:1242 weakly suppresses
  22. L9:1939 weakly excites
  23. L2:359 weakly excites
  24. L20:1300 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 -1.38 and -1.221 values between -1.07 and -0.911 values between -0.91 and -0.762 values between -0.76 and -0.601 values between -0.60 and -0.447 values between -0.44 and -0.29222 values between -0.29 and -0.132642 values between -0.13 and 0.03940 values between 0.03 and 0.1872 values between 0.18 and 0.347 values between 0.34 and 0.492 values between 0.80 and 0.961 values between 0.96 and 1.122 values between 1.12 and 1.272 values between 1.43 and 1.581 values between 2.05 and 2.211 values between 3.14 and 3.301 values between 3.77 and 3.921 values between 4.23 and 4.391 values between 4.70 and 4.86offnegativepositiveoften

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. ric
  2. ament
  3. (dot
  4. PEC
  5. *@
  6. 企业提供
  7. pery
  8. mark
  9. @}
  10. rama

Suppresses

  1. itta
  2. ao
  3. istes
  4. Endpoint
  5. ща
  6. uate
  7. itate
  8. ISTA
  9. 女主
  10. 畛