Neuron Explorer

Qwen3-0.6B

end of a clause before a verb or 'to'

A neuron in layer 22 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires when the text is about to continue with an action, like 'and' before 'type', 'to' before 'start', or 'using' after a tool name. When it was turned up, the model's writing got wandering and loose, with no clear theme.

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

    :⏎- Interpersonal, project management, risk management and time management skills.⏎- Software development in Java using various web-development technologies.⏎Figure 1: Traceability User Interface⏎Back to

  2. #2 peak

    ILEGES ON database_name.* TO user_name@localhost;⏎mysql> FLUSH PRIVILEGES;⏎Step 5: Installation and Configuration of Drupal⏎Open Browser and type “”http://Server-IP/drupal”” which will open below

  3. #3 peak

    format is very clear. This means that their non-compliance may be accompanied by fines and other losses on the side of the project executor.⏎Don’t know where to start?⏎We will advise you on the best way to realize your idea, leveraging

  4. #4 peak

    followed all the steps, you will see one shortcut on the desktop. Cut it and paste it at the following path⏎Now open Run (ctrl+R) and type your command name(Here my command name is “DOSBox“)

  5. #5 peak

    given to you?⏎October 3rd, 2010, 12:50 PM⏎i started the homework⏎i dont know how to finish it here what i

  6. #6 peak

    Select Debug >> Start without debugging (Ctrl+F5)⏎- For any runtime error, you can debug the program as follows:⏎- Select Debug >> Start debugging (F5)⏎- Please note that during debugging, the program will show and close the

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: when the "fog" in a "rural" or "farmed" village, from a local and informed and informed, "rural" and "farmed" in the "senses" of local, local and local

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

    same output.⏎- Nearly Every Database supports available⏎One of the greatest features in PHP is its support for a wide range of databases, nearly every databases – MySQL, ORACLE, PostgreSQL, Sybase, Informix to name a few.⏎What

  2. #2 peak

    ⏎- proper n. The northern parts of Scandinavia and Finland with the Kola Peninsula in Russia, traditionally inhabited by the Sami people.⏎from WordNet 3.0 Copyright 2006 by Princeton University

  3. #3 peak

    the server.⏎- PHP was created by Rasmus Lerdorf in 1995.⏎- Though PHP was originally known as Personal Home Page (PHP Tools), but, with the increase of its usefulness and

  4. #4 peak

    or plate current flows in an electron tube. In the case of a transistor, it is the electrical angle during which the collector or drain current flows. For a class A amplifier, the operating angle is 360°. That is,

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. L10:155 excites
  2. L7:92 excites
  3. L3:130 excites
  4. L2:24 suppresses
  5. L9:207 suppresses
  6. L15:184 excites
  7. L19:214 excites
  8. L21:158 excites
  9. L13:44 suppresses inside web addresses and citations
  10. L20:859 excites
  11. L8:328 excites
  12. L9:46 suppresses
  13. L4:342 weakly suppresses
  14. L4:92 weakly suppresses
  15. L11:840 weakly suppresses
  16. L4:76 weakly excites
  17. L11:223 weakly excites
  18. L4:42 weakly suppresses
  19. L5:7 weakly suppresses
  20. L10:556 weakly excites
  21. L19:400 weakly excites
  22. L10:1278 weakly suppresses
  23. L10:2230 weakly excites
  24. L18:1894 weakly excites

Feeds later layers

  1. L23:381 weakly excites
  2. L27:1219 weakly suppresses
  3. L23:2409 weakly suppresses
  4. L25:1838 weakly suppresses
  5. L26:687 weakly excites
  6. L25:2590 weakly suppresses
  7. L27:2775 weakly excites
  8. L24:2923 weakly suppresses
  9. L27:895 weakly excites
  10. L24:2987 weakly suppresses
  11. L24:223 weakly excites
  12. L24:738 weakly suppresses
  13. L24:154 weakly suppresses
  14. L25:1115 weakly excites
  15. L23:2911 weakly suppresses
  16. L23:377 weakly excites
  17. L23:2270 weakly excites
  18. L23:1881 weakly excites
  19. L26:1230 weakly excites
  20. L25:1330 weakly suppresses
  21. L24:2380 weakly excites
  22. L25:309 weakly excites
  23. L23:877 weakly suppresses
  24. L25:1149 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 -5.95 and -5.701 values between -3.44 and -3.192 values between -2.94 and -2.691 values between -2.69 and -2.444 values between -2.44 and -2.191 values between -2.19 and -1.942 values between -1.94 and -1.694 values between -1.44 and -1.193 values between -1.19 and -0.9410 values between -0.94 and -0.6952 values between -0.69 and -0.44258 values between -0.44 and -0.191070 values between -0.19 and 0.061543 values between 0.06 and 0.31691 values between 0.31 and 0.56184 values between 0.56 and 0.8152 values between 0.81 and 1.0618 values between 1.06 and 1.321 values between 1.32 and 1.572 values between 1.57 and 1.821 values between 1.82 and 2.072 values between 2.07 and 2.323 values between 2.32 and 2.571 values between 3.57 and 3.821 values between 3.82 and 4.07offnegativepositiveoften

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. 料
  2. 工具
  3. 方面
  4. ylation
  5. ić
  6. ноп
  7. ewise
  8. 开
  9. 那边
  10. 蕊

Suppresses

  1. 刍
  2. ropy
  3. Verg
  4. udi
  5. utable
  6. undler
  7. #+#
  8. ="'.
  9. anga
  10. anza