Neuron Explorer

Qwen3-0.6B

'the/as' before 'following'

A neuron in layer 26 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 words like 'the' or 'as' that come right before the word 'following', as in 'one of the following' or 'are as follows'. Turned down, the model kept repeating the word 'following'.

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

    lost when components are concatenated.⏎You can create a RAID 5 volume without having to “zero out” the data blocks. To do so, do one of the following:⏎Use the metainit command with the -k option. The -

  2. #2 peak

    of the business.⏎How dirty data happens⏎In Salesforce and most other CRMs, the origin of dirty data can usually be pinpointed by one or more of the following behaviors:⏎- Importing data from varying sources that lack proper structure or formatting

  3. #3 peak

    pay attention to its growing and implement various ways to care for it and make it grow. The various ways that we can implement to make Monstera grow are as follows

  4. #4 peak

    /files/projects/drupal-7.33.zip⏎Step 2: Extract Compressed Drupal⏎Extract the Drupal zip file in the /var/www/html directory using following command:⏎# cd /var/www/html⏎# unzip drupal-7

  5. #5 peak

    others, such as certain scans, may be ordered outside.⏎Once a diagnosis is made, a treatment plan is presented; this plan might include any or all of the

  6. #6 peak

    with mo additional cost.⏎(3) WHAT HAPPENS IN A MEDIATION -Mediation is a structured process and Comprises of various stages, which are as under :- 1. Introduction :- Mediation introduces the parties, explains mediation process

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: following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following following

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

    I am pleased to announce that gold processing operations have⏎Ore processing operations at Newmont Boddington Gold - AusIMM⏎The process flow sheet at Newmont B

  2. #2 peak

    ,” by Lt. Gen. Harold G. Moore and war correspondent Joseph Galloway. The 2002 movie, “We Were Soldiers,” starring Mel Gibson, was based on the Moore-Galloway book. According to Gallow

  3. #3 peak

    and famous relative Francis Scott Key—composer of the United States national anthem, "The Star Spangled Banner"—but was commonly known as 'Scott'. He spent

  4. #4 peak

    replacing or augmenting the electronicssil...⏎Introduceți subiectul și categoria pentru o căutare cât mai relevantă.<|endoftext|>How to be more confident in school

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. L21:1049 strongly suppresses
  2. L21:1816 strongly excites
  3. L25:1388 strongly excites
  4. L24:433 strongly excites
  5. L20:1679 suppresses
  6. L25:1729 excites
  7. L25:1414 excites
  8. L17:243 suppresses
  9. L21:2134 excites
  10. L18:2047 suppresses
  11. L18:2439 excites
  12. L18:2837 excites
  13. L18:2139 suppresses
  14. L16:2539 suppresses
  15. L25:702 suppresses
  16. L21:2426 suppresses
  17. L22:2156 suppresses
  18. L20:1077 excites
  19. L22:1607 excites
  20. L24:494 excites
  21. L16:1961 suppresses
  22. L24:945 excites
  23. L15:2158 suppresses
  24. L17:2313 weakly suppresses

Feeds later layers

  1. L27:2353 suppresses
  2. L27:239 weakly excites
  3. L27:697 weakly excites
  4. L27:753 weakly suppresses
  5. L27:2105 weakly suppresses
  6. L27:1099 weakly suppresses
  7. L27:690 weakly suppresses
  8. L27:1779 weakly suppresses
  9. L27:2471 weakly suppresses
  10. L27:2020 weakly excites
  11. L27:1711 weakly suppresses
  12. L27:1811 weakly excites
  13. L27:799 weakly excites
  14. L27:1107 weakly suppresses
  15. L27:2718 weakly suppresses
  16. L27:204 weakly excites
  17. L27:485 weakly excites
  18. L27:2393 weakly suppresses
  19. L27:1156 weakly suppresses
  20. L27:2994 weakly excites
  21. L27:2272 weakly suppresses
  22. L27:1148 weakly suppresses
  23. L27:2948 weakly suppresses
  24. L27:2637 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 -10.43 and -9.802 values between -7.91 and -7.291 values between -7.29 and -6.663 values between -4.77 and -4.144 values between -4.14 and -3.513 values between -3.51 and -2.887 values between -2.88 and -2.2524 values between -2.25 and -1.6334 values between -1.63 and -1.00154 values between -1.00 and -0.372528 values between -0.37 and 0.26991 values between 0.26 and 0.8997 values between 0.89 and 1.5225 values between 1.52 and 2.1513 values between 2.15 and 2.787 values between 2.78 and 3.411 values between 3.41 and 4.041 values between 4.04 and 4.664 values between 4.66 and 5.293 values between 6.55 and 7.181 values between 7.18 and 7.811 values between 7.81 and 8.441 values between 8.44 and 9.071 values between 9.70 and 10.321 values between 14.10 and 14.73offnegativepositiveoften

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. ่
  2. пас
  3. ño
  4. ur
  5. plit
  6. };⏎⏎⏎
  7. 又是
  8. };
  9. __(*
  10. pered

Suppresses

  1. 以下
  2. following
  3. 下列
  4. following
  5. 如下
  6. Following
  7. 以下是
  8. siguientes
  9. seguint
  10. 下面