Neuron Explorer

Qwen3-0.6B

connecting word in how-to steps

A neuron in layer 24 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 on the small word that links an action to its method in instructions, like 'by' in 'open IIS Manager by clicking' or 'then' in 'select everything then delete'. Turning it up made the model write in a plain, list-like explanatory style, though not consistently.

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

    easy to sort data). Turns out that it's very easy to sort data in MS-Word, here's how!⏎- Highlight the entire list.⏎- Go to the Table menu and choose the Sort option.⏎You'll see that the default is

  2. #2 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

  3. #3 peak

    website. The Host name should be entered.⏎How Do I Host A Simple Html Page In Iis?⏎You can select Internet Information Services (IIS) Manager by right clicking your server in Server Manager. You can add a website by right-click

  4. #4 peak

    the Outlook window, choose the transferred particulars folder.⏎- Then open the email that you want to recall. You must double-click to open the

  5. #5 peak

    7 and 8. You can open IIS Manager by clicking here. The Internet Information Services (IIS) Manager should now be open

  6. #6 peak

    >> Microsoft Visual C++ 2008 Express Edition >> Visual Studio Tools >> Visual Studio 2008 Command Prompt (Run with administrator privileges)⏎- Once you are in the command line, you can use any editor such as edit

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: What if we, as human beings, are very much in need of the four four four four four four four four four four four four four four four four four four four four four four four four four four four four four four four four four

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

    Child Well-Being Network (ICWBN) online workshop. The programme and recordings of which are available at⏎Read the full policy brief here.<|endoftext|>Click or tap here to find out

  2. #2 peak

    must use an ATM to request the funds⏎Another excellent example of multi-factor authentication is using an SMS, voice message, or email to send a code the user can input along with a password to gain access. In this case, you’re combining

  3. #3 peak

    zones: 3-10.⏎Heat zones: 8-1.⏎Height: 30-35 inches (76-88 cm).⏎Star-shaped, white petals recurved at the tips.⏎Humus

  4. #4 peak

    Atlantic oceans. The city’s elevation ranges from 5,700 to 5,930 feet (1,740 to 1,810 metres).⏎Aside from a few small streams and artificial lakes, Johannesburg

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. L23:1207 suppresses
  2. L23:3011 suppresses
  3. L15:1815 excites
  4. L21:2180 excites
  5. L23:1579 weakly suppresses
  6. L9:226 weakly suppresses 'to' starting an infinitive
  7. L20:1054 weakly suppresses
  8. L10:2528 weakly suppresses
  9. L19:2249 weakly excites
  10. L15:3000 weakly excites
  11. L13:2707 weakly suppresses
  12. L12:845 weakly excites
  13. L19:1757 weakly suppresses
  14. L20:1589 weakly suppresses
  15. L22:1524 weakly suppresses
  16. L9:2036 weakly suppresses
  17. L20:1474 weakly suppresses
  18. L19:730 weakly excites
  19. L0:2682 weakly excites
  20. L18:1163 weakly suppresses
  21. L2:305 weakly excites
  22. L1:2941 weakly suppresses
  23. L5:1288 weakly excites
  24. L9:2029 weakly suppresses

Feeds later layers

  1. L26:739 weakly suppresses
  2. L27:391 weakly suppresses
  3. L25:2519 weakly suppresses
  4. L26:2894 weakly suppresses
  5. L25:1478 weakly suppresses
  6. L25:2786 weakly suppresses
  7. L27:1673 weakly suppresses
  8. L26:2624 weakly excites
  9. L25:1464 weakly excites
  10. L25:333 weakly excites
  11. L27:2227 weakly suppresses
  12. L26:951 weakly excites
  13. L27:1418 weakly excites
  14. L26:2208 weakly suppresses
  15. L27:449 weakly excites
  16. L25:2298 weakly suppresses
  17. L27:2545 weakly excites
  18. L25:2427 weakly excites
  19. L27:486 weakly suppresses
  20. L25:1613 weakly excites
  21. L27:3 weakly suppresses
  22. L25:2285 weakly excites
  23. L26:1189 weakly excites
  24. L27:2888 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 -10.33 and -9.691 values between -9.69 and -9.051 values between -7.77 and -7.131 values between -7.13 and -6.491 values between -6.49 and -5.852 values between -5.21 and -4.583 values between -4.58 and -3.943 values between -3.94 and -3.305 values between -3.30 and -2.669 values between -2.66 and -2.0210 values between -2.02 and -1.3871 values between -1.38 and -0.741274 values between -0.74 and -0.102366 values between -0.10 and 0.54111 values between 0.54 and 1.1815 values between 1.18 and 1.829 values between 1.82 and 2.469 values between 2.46 and 3.105 values between 3.10 and 3.743 values between 3.74 and 4.383 values between 4.38 and 5.012 values between 6.29 and 6.931 values between 7.57 and 8.211 values between 8.85 and 9.491 values between 14.60 and 15.24offnegativepositiveoften

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. asta
  2. arend
  3. gress
  4. aje
  5. athing
  6. aste
  7. ogg
  8. edom
  9. ulative
  10. grades

Suppresses

  1. .scan
  2. 篦
  3. 扫描
  4. nib
  5. روس
  6. 扫
  7. iac
  8. rox
  9. òng
  10. haul