Neuron Explorer

Qwen3-0.6B

words before a required or needed object

A neuron in layer 8 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 words that sit right before the thing being needed, paid, or provided, such as 'if required', 'need it', or 'has to pay'. Turning it up gave no clear change in topic, so its role is hard to pin down.

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

    U Press / 2010 / Other$27.225 Stars Out Of 5 1 ReviewsAvailability: In StockCBD Stock No: WW265116⏎BJU Science Grade

  2. #2 peak

    such as the European Convention on Human Rights and its fourteen protocols and national laws such as the Freedom of Information Act 2000;⏎- seek advice as necessary to ensure that the institution does what is right both in the spirit and the letter

  3. #3 peak

    the MTSS system.⏎The purpose of this guide is to provide brief explanations of key practices that can be implemented when working with students in need of intensive intervention in mathematics. Special education instructors, math interventionists, and others

  4. #4 peak

    a mental health issue or emotional distress.⏎- They are not therapists or psychiatrists, but they can give you initial support and signpost you to appropriate help if required.⏎If you need help or have any questions about Mental Health First Aid please email

  5. #5 peak

    . They are soliciting suggestions from users on what they would like to see from Project Glass.⏎“We think technology should work for you—to be there when you need it and get out of your way when you don’t.⏎A group of us from

  6. #6 peak

    of the total cost of a healthcare service, while the insurance company covers the remaining portion.⏎e. Out-of-Pocket Maximum: The maximum amount individuals have to pay during a policy period, after which the insurance company covers all

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: "I need to write a report for a client, and I need to write a certain amount of work. The client has asked me to write a report, and I need to make sure that I don't have any other requirements. I

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

    English one, which was forced to disband due to the Reformation of 1534. The Maltese architect Gerolamo Cassar was commissioned with the design of the seven auberges.⏎Each

  2. #2 peak

    Soyuz spacecraft to ferry astronauts⏎to and from the International Space Station (ISS).⏎To address the United States’ commercial human spaceflight needs, the agency has contracted Space Exploration Technologies (SpaceX) to develop a launch escape system that enables the

  3. #3 peak

    for works that are sold for money. That is because registration helps to prove that the copyright of a work belongs to a certain author.⏎If an author gets paid to make a work for someone else, the person who pays for making the work (

  4. #4 peak

    , gestures, and dress, a result, in part, of looking to English sources for portraiture, but also a means of meeting the many requests for commissions. Beginning in 175

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. L6:1924 suppresses
  2. L6:2929 weakly suppresses
  3. L7:1887 weakly suppresses
  4. L6:2602 weakly suppresses
  5. L6:2333 weakly excites
  6. L3:531 weakly suppresses
  7. L6:1698 weakly excites
  8. L5:495 weakly suppresses
  9. L6:385 weakly suppresses
  10. L4:1943 weakly suppresses
  11. L0:384 weakly excites
  12. L4:772 weakly suppresses
  13. L7:2950 weakly excites
  14. L7:102 weakly suppresses
  15. L5:1090 weakly excites
  16. L2:1857 weakly suppresses
  17. L5:2110 weakly excites
  18. L5:102 weakly excites
  19. L4:1874 weakly suppresses
  20. L3:861 weakly excites
  21. L7:2753 weakly excites
  22. L4:3056 weakly suppresses
  23. L3:2648 weakly excites
  24. L7:113 weakly excites

Feeds later layers

  1. L10:2343 suppresses
  2. L10:2623 suppresses
  3. L18:2729 suppresses
  4. L10:1624 weakly suppresses
  5. L9:2093 weakly suppresses
  6. L22:2206 weakly suppresses
  7. L9:1140 weakly suppresses
  8. L12:1330 weakly suppresses
  9. L21:1619 weakly suppresses
  10. L9:1859 weakly suppresses
  11. L12:1817 weakly suppresses
  12. L12:1428 weakly excites
  13. L12:2187 weakly excites
  14. L9:2231 weakly excites
  15. L19:539 weakly suppresses
  16. L11:2915 weakly suppresses
  17. L17:1147 weakly excites
  18. L11:2907 weakly excites
  19. L10:2761 weakly suppresses
  20. L20:2845 weakly suppresses
  21. L11:1259 weakly suppresses coercion, abuse and harassment
  22. L9:639 weakly suppresses
  23. L18:2970 weakly excites
  24. L15:1040 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 -7.80 and -7.561 values between -3.72 and -3.482 values between -3.48 and -3.241 values between -2.28 and -2.041 values between -2.04 and -1.802 values between -1.56 and -1.322 values between -1.32 and -1.081 values between -1.08 and -0.847 values between -0.84 and -0.6063 values between -0.60 and -0.36759 values between -0.36 and -0.122423 values between -0.12 and 0.12605 values between 0.12 and 0.3636 values between 0.36 and 0.602 values between 0.60 and 0.841 values between 1.08 and 1.321 values between 1.56 and 1.80offnegativepositiveoften

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. avenport
  2. jang
  3. ki
  4. ding
  5. bs
  6. =logging
  7. pler
  8. Spo
  9. –⏎⏎
  10. Jer

Suppresses

  1. ushman
  2. assignable
  3. contracts
  4. ООО
  5. 该公司
  6. CONTRACT
  7. assign
  8. 下达
  9. ựa
  10. avors