Neuron Explorer

Qwen3-0.6B

'computer' and 'desktop' after device words

A neuron in layer 8 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the word 'computer' or 'desktop' when it follows phrases like 'to your' or 'through'. Turning it up did not give a clear theme in the output, so the link to the word is strong but the effect is not.

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

    ⏎Please Note: This ebook is a digital download, NOT a physical product. After purchase, you will be provided a one time link to download ebooks to your computer. Orders paid by PayPal require up to 8 business hours to verify payment and

  2. #2 peak

    be so confusing. Instead of purchasing an app on a disk at the physical location of a store, you download them from the Windows 8 Store directly to your computer.⏎Check out this article from our archives to learn more about apps: What Are

  3. #3 peak

    Canada and the Northern US. These satellites work by connecting to "earth stations" which can receive the satellite signals and channel the broadband internet to homes or businesses through fiber optic cable.⏎As SpaceX works to expand its capacity for more and more customers and

  4. #4 peak

    of “Fizzyx for Fun”, normally priced at $9.95, will be available for you to download for free, either on e-reader or desktop, until December 5 at https://www.smashwords.com/<|endoftext|>If

  5. #5 peak

    would design its own chips for Macs and MacBooks, moving away from Intel’s products. Apple leveraged the work it did on smartphone chips to develop its desktop ones, which

  6. #6 peak

    mobile technology. In augmented reality the line between the virtual world and the real one is blurred. Engineers use some sophisticated technologies to pull out graphics from television screen and computer display and integrate them into real world environments. Though it is

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 have been in the role of a position, and in the security of having access to information, the ability to maintain confidentiality of the information, and the other security. In addition, the ability to access to the management of information to the

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

    Mobile App: Apple Store | Google Play⏎Dear visitor, the comments section is undergoing an overhaul and will return soon.

  2. #2 peak

    phones and/or data.⏎Some local statistics:⏎- Smart phone penetration around 51%⏎- >80% use cellphones to communicate⏎- 4G/LTE subscriptions 22%⏎- 4G/LTE coverage 8

  3. #3 peak

    24th October, 1946, when a V-2 rocket lifted the first ever space-bound camera above New Mexico on a short sub-orbital flight. From an altitude of 65 miles, a 35

  4. #4 peak

    propagation mechanisms⏎Radio propagation modelling⏎Production of radio path profiles⏎Calculating power budgets⏎Calculating fade margins⏎Planning coverage and capacity for mobile systems⏎Radio site engineering⏎Design a mobile radio system⏎Design a fixed radio

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. L4:310 suppresses
  2. L6:1362 excites
  3. L7:2920 excites
  4. L7:3026 weakly suppresses
  5. L7:536 weakly suppresses
  6. L3:1639 weakly excites
  7. L6:2629 weakly suppresses
  8. L6:1781 weakly suppresses
  9. L5:28 weakly excites
  10. L7:659 weakly excites
  11. L6:1030 weakly suppresses
  12. L6:6 weakly excites
  13. L4:840 weakly suppresses
  14. L1:2020 weakly excites
  15. L7:896 weakly excites
  16. L6:2677 weakly excites
  17. L5:135 weakly suppresses
  18. L3:1002 weakly excites
  19. L7:2341 weakly excites
  20. L3:249 weakly excites
  21. L7:2216 weakly excites
  22. L3:334 weakly suppresses
  23. L3:1405 weakly excites
  24. L1:739 weakly excites

Feeds later layers

  1. L17:1074 strongly suppresses
  2. L20:1625 strongly suppresses
  3. L9:502 strongly suppresses
  4. L23:2084 suppresses
  5. L21:1543 suppresses
  6. L11:1148 excites
  7. L23:3028 suppresses
  8. L25:1617 suppresses
  9. L26:283 suppresses
  10. L20:1132 suppresses
  11. L20:2619 weakly suppresses
  12. L10:997 weakly suppresses
  13. L10:70 weakly suppresses
  14. L27:260 weakly suppresses
  15. L23:1735 weakly excites
  16. L16:1470 weakly suppresses
  17. L23:3008 weakly suppresses
  18. L20:1209 weakly excites
  19. L25:1217 weakly excites
  20. L21:949 weakly suppresses
  21. L17:2376 weakly suppresses
  22. L20:1676 weakly suppresses
  23. L21:575 weakly excites
  24. L9:1832 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 -1.55 and -1.411 values between -1.27 and -1.131 values between -1.13 and -0.991 values between -0.99 and -0.854 values between -0.71 and -0.5717 values between -0.57 and -0.4387 values between -0.43 and -0.29394 values between -0.29 and -0.141352 values between -0.14 and -0.001485 values between -0.00 and 0.14451 values between 0.14 and 0.2880 values between 0.28 and 0.4212 values between 0.42 and 0.566 values between 0.56 and 0.704 values between 0.70 and 0.843 values between 0.98 and 1.121 values between 1.12 and 1.261 values between 1.26 and 1.401 values between 1.54 and 1.681 values between 1.82 and 1.962 values between 2.10 and 2.242 values between 2.52 and 2.661 values between 3.93 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. Knot
  2. licate
  3. steen
  4. tableFuture
  5. 蚀
  6. Princeton
  7. fibers
  8. Principles
  9. SpaceItem
  10. nin

Suppresses

  1. _android
  2. Android
  3. 安全感
  4. Android
  5. android
  6. 安卓
  7. phone
  8. android
  9. phone
  10. телеф