Neuron Explorer

Qwen3-0.6B

the word 'very'

A neuron in layer 17 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 lights up on the word 'very', especially in phrases like 'The Very Basics' or 'very cool' and 'very own'. When it was turned up, the model's writing kept coming back to 'very' ('very different', 'very cold'). Turned down, it tended to repeat 'every' instead.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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

    in your head) is vital so that you have a timing grid where you can follow the notes as they pass. Refer to the sheet ‘Reading Music: The Very Basics - Examples’ for some simple lines to get all this information working for you

  2. #2 peak

    ave, b) find them on the bass and c) recognise different rhythms. Watch this lesson first for a recap: Reading Music On The Bass Guitar: The Very Basics - Note Names And Locations.⏎Each one of those things is quite easy to

  3. #3 peak

    the Web: a short history⏎Before we get onto more practical things, let’s take a look at how animation on the Internet came to its present (and very cool) state. It pretty much all started with gifs…⏎.gif files are,

  4. #4 peak

    hidden by what we see." --Rene Magritte⏎D.B. Johnson writes and illustrates the surreal story of famous surrealist painter Rene Magritte and his very mysterious (and mischievous ) hat. While the art reflects some of Mag

  5. #5 peak

    to be a widespread phenomenon in mammals, however, and you don’t need to (and shouldn’t!) take my word for it – the following is from a very recent review by leaders in this field:⏎"...epigenetic inheritance is usually—if

  6. #6 peak

    and philosophical principles, this is a thorough exploration of the concept and practice of⏎liberalism.ABOUT THE SERIES: The Very Short Introductions series from Oxford University Press contains hundreds of titles in almost every

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: Every year, the number of each of the 365 days each day each year is every 365 times each 365 each, and so each day, every 365 each each, every each,

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

    pigment that produces color in other breeds of turkey. While they may not be as common as the more traditional dark-feathered birds, white turkeys can be found at specialty farms across the country. If you’re lucky enough to come across one

  2. #2 peak

    enjoy and can do daily, such as walking or jogging, swimming, cycling, high-intensity interval training, or even dancing. Also include strength training exercise at least two days per week.⏎Stress can elevate blood pressure too. Find ways to

  3. #3 peak

    , millions of people still cannot access it: just 32% of children who are HIV-positive are currently on treatment in sub-Saharan Africa.. Every day at least 400 children have new HIV Infections in Southern Africa⏎The HIV

  4. #4 peak

    This results in steadier blood sugar levels and sustained energy. Also, complex carbohydrates tend to provide more vitamins and minerals than sources of simple sugars.⏎Sugar has its place in your diet, but many health and diet experts warn that excess sugar consumption can

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. L4:1873 suppresses
  2. L1:1158 weakly excites
  3. L13:26 weakly excites
  4. L14:2713 weakly excites
  5. L12:1819 weakly suppresses
  6. L16:2643 weakly suppresses
  7. L1:1361 weakly suppresses
  8. L16:88 weakly excites
  9. L2:1863 weakly excites
  10. L5:1908 weakly suppresses
  11. L15:542 weakly suppresses
  12. L12:503 weakly suppresses
  13. L15:1614 weakly excites
  14. L1:2464 weakly excites
  15. L11:2650 weakly excites
  16. L4:2160 weakly suppresses
  17. L13:2019 weakly excites
  18. L2:316 weakly excites
  19. L16:574 weakly excites
  20. L15:1515 weakly suppresses
  21. L15:2473 weakly excites
  22. L5:2927 weakly excites
  23. L0:3012 weakly excites
  24. L15:12 weakly suppresses

Feeds later layers

  1. L20:2464 suppresses
  2. L27:1187 weakly suppresses
  3. L26:890 weakly suppresses
  4. L20:2962 weakly excites "of" in "most X of all" superlatives
  5. L19:2503 weakly suppresses
  6. L19:2868 weakly suppresses
  7. L22:1033 weakly suppresses
  8. L24:1149 weakly suppresses
  9. L19:949 weakly excites
  10. L27:2910 weakly suppresses
  11. L24:1179 weakly excites
  12. L24:1282 weakly excites
  13. L19:1063 weakly excites
  14. L25:970 weakly excites
  15. L20:1012 weakly excites
  16. L25:596 weakly suppresses words right before 'all' or 'in'
  17. L20:2076 weakly suppresses
  18. L23:97 weakly suppresses
  19. L18:1774 weakly excites
  20. L22:1998 weakly suppresses
  21. L23:1305 weakly suppresses
  22. L18:349 weakly suppresses
  23. L20:2774 weakly suppresses
  24. L21:1352 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 -4.88 and -4.441 values between -4.44 and -4.001 values between -3.56 and -3.124 values between -3.12 and -2.698 values between -2.69 and -2.259 values between -2.25 and -1.8123 values between -1.81 and -1.3746 values between -1.37 and -0.93134 values between -0.93 and -0.49978 values between -0.49 and -0.052294 values between -0.05 and 0.39244 values between 0.39 and 0.8369 values between 0.83 and 1.2738 values between 1.27 and 1.7123 values between 1.71 and 2.1511 values between 2.15 and 2.598 values between 2.59 and 3.035 values between 3.03 and 3.472 values between 3.47 and 3.911 values between 3.91 and 4.351 values between 4.35 and 4.791 values between 5.23 and 5.671 values between 5.67 and 6.111 values between 6.11 and 6.551 values between 6.55 and 6.991 values between 7.87 and 8.311 values between 12.26 and 12.70offnegativepositiveoften

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. otion
  2. 他自己
  3. 看他
  4. himself
  5. nghề
  6. -ending
  7. themselves
  8. 蹊
  9. 其他人
  10. 别的

Suppresses

  1. 每一次
  2. each
  3. 每个
  4. Each
  5. 每个月
  6. 每一
  7. Each
  8. 每位
  9. 每周
  10. annually