Neuron Explorer

Qwen3-0.6B

the word 'following' after 'the'

A neuron in layer 2 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 'following' when it comes right after 'the', as in 'the following modules' or 'the following data'. It almost never fires on anything else. Turning it up did not produce a visible change in the model's text.

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

    barriers to trade (TBT), trade-related aspects of intellectual property rights (TRIPS), and trade in services.⏎Content and Structure⏎The course consists of the following modules:⏎• Module I – The Concept of Trade Negotiations: An Overview⏎

  2. #2 peak

    blog post we will use a Python script to work out how many lego bricks would be needed to build planet Earth!⏎To solve our problem we will use the following data:⏎We will work out the number of lego bricks needed in several step

  3. #3 peak

    . These protections are designed to ensure equal opportunities and fair treatment in employment and other areas of life.⏎The protected classes of individuals in California labor law typically include the following:⏎Race and Ethnicity⏎Individuals are protected from discrimination based on their race

  4. #4 peak

    can be treated with Cyberknife?⏎The Cyberknife System is designed for radiosurgical treatment of tumors and lesions throughout the body. The current range of therapies includes the following diseases of the head: acoustic neuromas / vestibular schwannomas,

  5. #5 peak

    . Flame rods are normally used for detecting flames from gas. Oil generated flames are not eligible for the application of the flame rods. The flame rods comes with the following advantages.⏎• It provides a quick response to the failure of the flame.⏎•

  6. #6 peak

    years to maturity; and⏎YTM– The yield to maturity.⏎Let's take Bond A, issued by Company Alpha, as an example. It has the following data:⏎- Face value: $1,000⏎- Annual coupon

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 lately, and I have been thinking about lately, I have been thinking about lately, I have been thinking about lately, I have been thinking about lately... and I have been thinking about lately... and I have been

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

    like people do. That’s what a new study from an Austrian university revealed after they observed dogs giving food to dogs

  2. #2 peak

    in length (does not include title and reference page)⏎1-2 outside sources⏎Why Work with CheggAnswers⏎At CheggAnswers.com, we boast of our quick turnaround that is facilitated by our experienced team of writers. We can handle

  3. #3 peak

    the alphabet. I later assessed them on their alphabet knowledge and saw improvement from most, if not all, of the children.⏎What advice would you give?⏎Find what you love to do. Than your work will not seem like work at all.

  4. #4 peak

    term disability including brain and spinal cord injuries for ages 1 – 44.⏎As a society, driving is the only thing we do that makes us all the same. Motor vehicles are the true equalizer. Remember driving is a privilege not

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. L1:462 weakly excites
  2. L1:1645 weakly excites
  3. L1:959 weakly excites
  4. L0:1848 weakly excites
  5. L1:2150 weakly suppresses
  6. L0:918 weakly excites
  7. L1:136 weakly suppresses
  8. L1:2156 weakly suppresses
  9. L0:1052 weakly excites
  10. L1:2557 weakly excites
  11. L0:424 weakly suppresses open parenthesis after a name or term
  12. L1:1029 weakly excites
  13. L1:2661 weakly excites
  14. L0:1082 weakly suppresses
  15. L0:474 weakly excites
  16. L0:376 weakly excites
  17. L0:1845 weakly suppresses
  18. L0:1458 weakly suppresses
  19. L0:1456 weakly excites
  20. L1:2683 weakly excites
  21. L1:2533 weakly suppresses
  22. L1:1081 weakly excites
  23. L1:2980 weakly suppresses
  24. L0:606 weakly suppresses

Feeds later layers

  1. L19:2448 weakly suppresses
  2. L9:3067 weakly excites
  3. L16:651 weakly excites
  4. L5:903 weakly suppresses
  5. L20:1652 weakly suppresses
  6. L16:1258 weakly suppresses
  7. L20:678 weakly excites
  8. L7:2713 weakly excites
  9. L18:1406 weakly suppresses
  10. L25:2930 weakly suppresses
  11. L12:1929 weakly suppresses
  12. L22:334 weakly excites
  13. L13:626 weakly excites
  14. L25:762 weakly excites
  15. L13:508 weakly excites
  16. L11:2821 weakly suppresses
  17. L15:1035 weakly excites
  18. L9:1371 weakly excites
  19. L10:2877 weakly suppresses
  20. L26:2993 weakly excites
  21. L17:1780 weakly suppresses
  22. L17:2711 weakly suppresses
  23. L6:1338 weakly suppresses
  24. L4:2110 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 -1.79 and -1.606 values between -1.02 and -0.838 values between -0.83 and -0.6321 values between -0.63 and -0.4458 values between -0.44 and -0.25717 values between -0.25 and -0.052619 values between -0.05 and 0.14420 values between 0.14 and 0.3333 values between 0.33 and 0.537 values between 0.53 and 0.727 values between 0.72 and 0.913 values between 0.91 and 1.112 values between 1.30 and 1.492 values between 1.49 and 1.691 values between 2.84 and 3.041 values between 3.23 and 3.421 values between 5.74 and 5.93offnegativepositiveoften

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. lucent
  2. ointment
  3. 老人家
  4. Pocket
  5. 洑
  6. pomys
  7. fold
  8. cloudy
  9. erture
  10. Pocket

Suppresses

  1. ored
  2. ian
  3. '</
  4. orian
  5. ')</
  6. случая
  7. asto
  8. га
  9. .'</
  10. 钊