Neuron Explorer

Qwen3-0.6B

end-of-text right after cut-off text

A neuron in layer 20 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 invisible end-of-document marker that appears right after a sentence stops in the middle, like 'surgeons may work late-night or early-morning' followed by a break. Turning it up did not make the text clearly about endings, so the link to the steering output is unclear.

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

    outside, but when the wind was blowing from the South Pole plateau, it felt a lot colder! You have to wrap up in many layers to stay warm –<|endoftext|>Why Gas Masks Are a Necessity for our Society⏎We live in very

  2. #2 peak

    from the fenced perimeter, be prepared to possibly be stopped and answer their questions regarding your visit. The staff housing located not far from the main buildings are not secluded<|endoftext|>Do you know about Menstrual Hygiene Day, which is observed annually on

  3. #3 peak

    from intense red or yellow to the golden or brown hue. Such meticulous care means that the dates you will buy in handy package at the store is rich in natural<|endoftext|>If you are conscious about your teeth, you must know how to take care of

  4. #4 peak

    caught sight of a pale little girl with a thin neck, and spindling legs half-hidden in the folds of Aunt Frances’s skirts. But she didn’t<|endoftext|>Dec 6, 2023⏎Airbus has successfully passed the

  5. #5 peak

    resistant” may be a special treat to hungry deer, as they munch on tender soft new growth. Like many other animals, deer have food they avoid, and<|endoftext|>Seven miles to the south west of Sheffield there's a piece of wild moor

  6. #6 peak

    may be on call one day of the week or all week. Surgeons may work late-night or early-morning<|endoftext|>WK 4 SOCW 6443 Discussion 2: Psychoph

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: with the first tailed part of the main part of the main part of the main part of the main part of the main part, it's part of the main part. Wait, no, don't get the thing. Let's just

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

    table with a specific name or index exists.⏎The following example retrieves a single table by name from the DataTableCollection.⏎Private Sub GetTableByName() ' Presuming a DataGrid is displaying more than one table

  2. #2 peak

    Clause in oracle. In following section i would like to explain about the syntax as well as multiple examples of with clause.⏎With SQL_Query_Name As⏎Select * from New_SQL_Query_name;⏎Execution of With Clause :⏎In this section i would like

  3. #3 peak

    Implementation⏎You can use OpenCV to manipulate the images.⏎pip3 install opencv-python numpy⏎Open up a new Python file and follow along:⏎import numpy as np⏎As you have to follow a few steps, I would suggest you go

  4. #4 peak

    2: Extract Compressed Drupal⏎Extract the Drupal zip file in the /var/www/html directory using following command:⏎# cd /var/www/html⏎# unzip drupal-7.33.zip⏎It will create a folder as /

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. L19:1616 suppresses
  2. L8:446 suppresses
  3. L19:2516 excites
  4. L16:2984 excites
  5. L19:1416 excites
  6. L19:2136 suppresses
  7. L19:2672 suppresses
  8. L15:1434 excites
  9. L10:1237 excites
  10. L19:3032 suppresses
  11. L12:1966 excites
  12. L16:2165 suppresses
  13. L17:1532 excites
  14. L9:1657 excites
  15. L13:1466 excites
  16. L7:537 excites period in a middle initial
  17. L17:1158 suppresses
  18. L17:1203 excites
  19. L13:2608 suppresses
  20. L17:1236 suppresses
  21. L18:932 excites
  22. L18:2040 suppresses
  23. L5:666 suppresses
  24. L9:1424 suppresses

Feeds later layers

  1. L21:7 excites mixed, weak code and reference hits
  2. L25:460 suppresses
  3. L21:105 excites
  4. L26:168 suppresses
  5. L25:5 suppresses
  6. L21:142 weakly suppresses
  7. L24:85 weakly suppresses
  8. L21:2070 weakly excites
  9. L25:94 weakly suppresses
  10. L21:838 weakly excites
  11. L23:354 weakly suppresses
  12. L25:2198 weakly excites
  13. L24:686 weakly excites
  14. L21:166 weakly suppresses
  15. L23:3 weakly suppresses
  16. L22:640 weakly suppresses
  17. L25:362 weakly suppresses
  18. L24:1043 weakly excites
  19. L21:696 weakly suppresses
  20. L26:69 weakly suppresses
  21. L21:1227 weakly suppresses
  22. L22:29 weakly suppresses
  23. L21:238 weakly suppresses
  24. L21:115 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 -6.59 and -6.221 values between -5.14 and -4.781 values between -2.60 and -2.241 values between -2.24 and -1.882 values between -1.88 and -1.525 values between -1.52 and -1.169 values between -1.16 and -0.7982 values between -0.79 and -0.431021 values between -0.43 and -0.072420 values between -0.07 and 0.29309 values between 0.29 and 0.6540 values between 0.65 and 1.023 values between 1.02 and 1.384 values between 1.38 and 1.743 values between 1.74 and 2.101 values between 2.10 and 2.461 values between 2.46 and 2.831 values between 3.55 and 3.911 values between 6.08 and 6.451 values between 6.45 and 6.811 values between 7.53 and 7.90offnegativepositiveoften

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. forthcoming
  2. ปา
  3. 袱
  4. urrence
  5. istribution
  6. ocket
  7. istributions
  8. 告诉她
  9. unpack
  10. 诰

Suppresses

  1. ilated
  2. plate
  3. avn
  4. ivities
  5. main
  6. Tester
  7. enced
  8. ivity
  9. tester
  10. 五星