Neuron Explorer

Qwen3-0.6B

closures and construction

A neuron in layer 7 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 about things being closed or under construction, such as 'close', 'closed', 'closing' and 'construction'. Turning it up made the model write about closing a website and shutting things down, which matches the trigger.

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

    56 low-end items can be traded, like tea, bicycles and canned food. And for about half the year, heavy snowfall forces authorities to close the border altogether.⏎The connectivity problems in the Himalayas, long a natural land

  2. #2 peak

    published our first post on IDEA Public Schools we received LOTS of additional questions. We hope this post answers some of those questions!⏎IDEA Public Schools, Coming Soon to Southern Louisiana, is a Public Charter School Network—What Does That Mean?⏎

  3. #3 peak

    itage site, Science centre, Garden, parklands or rural site⏎Daily 12 noon to 5pm April to September only⏎Fossil Grove is closed from October to April⏎Parking for visitors is on the main road running adjacent to

  4. #4 peak

    to enjoy our nation’s natural heritage for generations to come.⏎Announcement: We are moving to a new web site. There may be some disruption of service from 3/23/13-3/24/13<|endoftext|>

  5. #5 peak

    would get worse when the Blue Grass Parkway was completed in the mid-1960s.⏎“Part of the road was raised and other things happened during construction which made the flooding worse,” Finney said.⏎On top of that, the

  6. #6 peak

    people the power to control an audio environment.⏎In public spaces, there are many noise pollutions that one doesn’t want to listen to such as traffic noise or construction noise. Digital devices like headphones or earbuds help people calm down by escaping

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: the idea of a group of people, and the idea of a body. So, when you have a situation where there is an object, and the idea of a body, and the idea of a person, and the idea of a person

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

    that the money raised by lotteries is used for good causes in the public sector, and it may reduce overall taxation.⏎A typical lottery involves players purchasing a ticket for a certain amount of money, which is then entered into a draw to win

  2. #2 peak

    is in no way intended to replace human translation. Neither BioOne nor the owners and publishers of the content make, and they explicitly disclaim, any express or implied representations or warranties of any kind, including, without limitation, representations and warranties as

  3. #3 peak

    .⏎Symptoms of early and late stage asthma⏎Normally on inhalation of allergens, the airway narrows and wheezing develops within minutes as the lungs swell and passageways narrow. The wheezing will start in about

  4. #4 peak

    some cases, a percentage of the total pool is reserved for costs related to organizing and promoting the lottery.⏎The first step in winning the lottery is to purchase a ticket, which can be done online or at any store that sells state-approved tickets.

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. L6:1298 suppresses
  2. L6:2796 excites
  3. L6:2345 excites
  4. L6:1965 weakly excites
  5. L6:3030 weakly excites
  6. L5:771 weakly excites
  7. L3:2772 weakly excites
  8. L6:2540 weakly suppresses
  9. L5:267 weakly suppresses
  10. L6:1355 weakly excites
  11. L6:1966 weakly suppresses
  12. L6:1842 weakly suppresses
  13. L0:566 weakly excites
  14. L6:12 weakly excites
  15. L2:2875 weakly suppresses 'up' after show, turn and showed
  16. L5:380 weakly suppresses
  17. L5:788 weakly excites
  18. L6:890 weakly suppresses
  19. L3:1135 weakly excites
  20. L6:1721 weakly excites
  21. L2:1899 weakly excites
  22. L5:1527 weakly excites
  23. L4:1034 weakly excites
  24. L6:2898 weakly suppresses

Feeds later layers

  1. L9:2909 excites
  2. L9:915 excites
  3. L10:2139 excites
  4. L11:587 excites
  5. L8:2646 excites
  6. L11:1076 excites
  7. L8:703 excites
  8. L15:1672 weakly excites
  9. L8:2890 weakly excites
  10. L16:1747 weakly excites
  11. L12:2185 weakly excites
  12. L11:2548 weakly excites
  13. L9:409 weakly suppresses
  14. L11:418 weakly excites unclear: scattered mid-phrase word pieces
  15. L11:2317 weakly excites
  16. L12:1774 weakly excites
  17. L8:1711 weakly excites
  18. L10:2444 weakly excites
  19. L14:1164 weakly suppresses
  20. L10:728 weakly suppresses
  21. L12:2377 weakly excites
  22. L11:1636 weakly excites
  23. L19:2725 weakly excites
  24. L8:721 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 -2.19 and -2.091 values between -1.30 and -1.201 values between -0.90 and -0.802 values between -0.80 and -0.714 values between -0.61 and -0.5114 values between -0.51 and -0.4138 values between -0.41 and -0.31157 values between -0.31 and -0.21511 values between -0.21 and -0.11987 values between -0.11 and -0.011234 values between -0.01 and 0.09612 values between 0.09 and 0.19236 values between 0.19 and 0.2975 values between 0.29 and 0.3920 values between 0.39 and 0.495 values between 0.49 and 0.594 values between 0.59 and 0.681 values between 0.68 and 0.781 values between 1.08 and 1.181 values between 1.28 and 1.381 values between 1.58 and 1.682 values between 1.68 and 1.78offnegativepositiveoften

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. 停电
  2. rendering
  3. 舂
  4. 不得转载
  5. interruption
  6. Bugün
  7. shifting
  8. Observatory
  9. ʕ
  10. modifier

Suppresses

  1. urus
  2. trinsic
  3. 材
  4. ets
  5. warts
  6. ys
  7. rama
  8. _complete
  9. ova
  10. alysis