Neuron Explorer

Qwen3-0.6B

'of' after 'benefits' or 'features'

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 fires on the word 'of' when it follows a phrase like 'benefits', 'main features' or 'basic principles', right before the topic named. Turning it up gave text that does not clearly carry this pattern.

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

    code, and automatically creates, changes, and manages the defined resources by parsing and executing the template to reach the goal of automated O&M.⏎Learn the main features of Terraform, you can refer to Getting Started with Terraform with Alibaba Cloud:

  2. #2 peak

    you will be familiar with the relevant documents on measurements and calibrations and will be able to determine measurement uncertainty to perform measurement tasks.⏎You learn the basic principles of evaluating measurement uncertainty according to

  3. #3 peak

    in coffee strengthen body organs, useful against inflammation and uplift our muscles and skin. People who consume coffee regularly are expected to live a long life. Six amazing benefits of coffee are given below: –⏎- Coffee helps reduce depression and make you happier

  4. #4 peak

    their back, especially a few weeks after being born. If they’re not lying on their backs, they are being carried or swaddled. A few benefits of tummy time include strengthening of neck, back, and upper body muscles.⏎Sp

  5. #5 peak

    and fully benefit from their participation by becoming equal partners in the negotiating process. The course will provide a comprehensive overview of the concept of negotiations, background and special characteristics of multilateral trade negotiations in

  6. #6 peak

    loads will be completed in the cloud.⏎Cloud computing initiates a grand shift from the old traditional IT management systems to new and improved digitally-powered administration. Some benefits of cloud computing include:⏎- Reduction of capital costs⏎- Improved functioning speed⏎-

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 know I have been thinking about what to do with this. I know, I feel confident that I can answer now with something. If I need to, I can answer within 10 minutes. So, what if I thought 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

    how to develop a design. We will also explain in details the important phase of making construction document. Then finally we will be actually designing a home using an extremely easy yet powerful and free 3Ds program

  2. #2 peak

    000 calorie diet. Please remember this when using this information to make healthy food choices for your diet.⏎Calories -

  3. #3 peak

    the following properties or attributes:⏎- Expiration Date⏎- Symbol or Underlying⏎- Strike Price⏎- Type: Call or Put⏎The contract has an expiration date. This is how long the option contract lives.⏎The contract also has an

  4. #4 peak

    not only mathematicians - like to consider more specific questions in order to get some feeling for what is happening. In this article we will reflect about a special class of Lunda-designs.⏎On the day before my daughter Likilisa's fourth

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. L6:1672 excites
  2. L5:1150 weakly excites
  3. L6:313 weakly excites
  4. L0:428 weakly suppresses
  5. L4:538 weakly suppresses
  6. L6:2585 weakly excites
  7. L6:1683 weakly excites
  8. L5:1384 weakly suppresses
  9. L6:1571 weakly excites
  10. L5:1131 weakly suppresses
  11. L6:397 weakly excites
  12. L5:439 weakly suppresses
  13. L5:1252 weakly excites
  14. L5:1556 weakly excites
  15. L4:1018 weakly suppresses
  16. L5:75 weakly excites
  17. L6:240 weakly suppresses
  18. L3:1709 weakly suppresses
  19. L1:700 weakly suppresses
  20. L1:1407 weakly suppresses
  21. L5:2849 weakly suppresses
  22. L4:2801 weakly excites
  23. L2:1973 weakly excites
  24. L5:2181 weakly suppresses

Feeds later layers

  1. L8:1179 excites
  2. L13:2498 excites
  3. L17:2548 weakly excites
  4. L12:2543 weakly excites
  5. L20:1324 weakly excites
  6. L10:2128 weakly excites
  7. L12:2892 weakly excites
  8. L13:2081 weakly excites
  9. L19:2212 weakly excites "what" after verbs of knowing
  10. L9:1474 weakly excites
  11. L9:1441 weakly excites
  12. L20:1016 weakly excites
  13. L9:488 weakly excites
  14. L11:1650 weakly excites
  15. L9:1852 weakly excites
  16. L20:2920 weakly excites
  17. L10:1174 weakly suppresses
  18. L22:1900 weakly excites
  19. L8:323 weakly excites
  20. L8:133 weakly excites
  21. L18:1815 weakly suppresses
  22. L26:463 weakly excites digits inside years and numbers
  23. L13:1263 weakly suppresses
  24. L17:2438 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 -2.10 and -1.931 values between -1.60 and -1.441 values between -1.44 and -1.271 values between -1.27 and -1.114 values between -1.11 and -0.957 values between -0.95 and -0.7815 values between -0.78 and -0.6230 values between -0.62 and -0.4562 values between -0.45 and -0.29274 values between -0.29 and -0.122074 values between -0.12 and 0.041082 values between 0.04 and 0.21257 values between 0.21 and 0.3756 values between 0.37 and 0.5414 values between 0.54 and 0.707 values between 0.70 and 0.862 values between 0.86 and 1.032 values between 1.03 and 1.192 values between 1.19 and 1.362 values between 1.36 and 1.522 values between 1.52 and 1.691 values between 1.69 and 1.851 values between 1.85 and 2.011 values between 2.01 and 2.183 values between 2.18 and 2.342 values between 2.34 and 2.511 values between 2.51 and 2.671 values between 3.00 and 3.171 values between 3.17 and 3.331 values between 4.32 and 4.48offnegativepositiveoften

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. oeff
  2. 刍
  3. briefly
  4. <body
  5. 옹
  6. rch
  7. los
  8. ồn
  9. izador
  10. во

Suppresses

  1. marginTop
  2. paddingHorizontal
  3. 族自治
  4. xl
  5. hành
  6. }'.
  7. jest
  8. Accountability
  9. _reporting
  10. ;');⏎