Neuron Explorer

Qwen3-0.6B

word after help, support or harm

A neuron in layer 16 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 responds to a word that comes right after a phrase about help, support, injury, or being targeted, such as "with the help," "specifically advised," or "got bit." It is hard to name a single theme, and steering did not make the model talk about it clearly.

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

    a frequent check list of possible symptoms, on a daily or weekly interval.⏎Do NOT be on constant vigilance for all possible cancer symptoms unless you have been specifically advised to do so by your oncology team.⏎Eternal Vigilance for months

  2. #2 peak

    these same issues. In 2012, the Islamic Center of Murfreesboro, Tennessee won the right to move into its new mosque with the help of a RLUIPA suit brought by the Department of Justice

  3. #3 peak

    , learning about their struggles and telling them they weren’t alone, an idea for a local support group emerged. Encouraged by the support of her supervisor and funding provided by a partial community grant, Dr. Kinlen engaged the families she coun

  4. #4 peak

    of performance. The techniques that have been described so far are helpful in increasing athletes’ short-term motivation.<|endoftext|>Is this insect bite making me sick?⏎I got bit by a strange insect yesterday. Didn't think much of it but today it is

  5. #5 peak

    Other signs of anxiety such as obsessive ideation and rumination can interfere with concentration and the capacity to process information efficiently. Social anxiety, including concerns about the judgment of peers, can cause kids to become overly cautious and slow in completing tasks. Sleeping

  6. #6 peak

    Library in Omaha, Nebraska, looks more like some sort of rodent than an impactful part of history. However, Thompson’s story—surviving a scalping, holding on to the coiffure in question—makes it all the more

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 a 3D model of a 3D model of a 3D model of a 3D model of a 3D model of a 3D model of a 3D model of a 3D

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

    or, this person freely and uncontrollably expresses their anger. They take no prisoners and have no mercy in the heat of the moment. Anyone and everyone can be a target for their verbal or physical expressions of anger.

  2. #2 peak

    Links: tips, tutorials, scripts, and more.⏎- Have an Oracle or SQL tip to offer your fellow DBAs and developers? The best tips submitted will receive a cool prize. Submit your tip today!⏎- Ask your technical Oracle and SQL

  3. #3 peak

    do is have a conversation with your tween about money. Speaking to them one-on-one is a big deal. The more you talk about money now, the better they will be off in the future. Remember that it’s never too late or too

  4. #4 peak

    mental health. When individuals focus on their work and lose perspective of those around them, they are also less likely to identify fellow members of the MIT community that may be susceptible to

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. L14:1973 strongly excites
  2. L11:1964 excites
  3. L12:2565 excites
  4. L15:3067 excites
  5. L13:1915 suppresses
  6. L11:915 suppresses
  7. L8:389 suppresses
  8. L7:1813 excites
  9. L12:953 excites
  10. L14:1647 excites
  11. L11:117 suppresses
  12. L12:2651 suppresses
  13. L10:204 excites
  14. L13:2052 excites
  15. L12:2634 suppresses
  16. L8:1005 suppresses
  17. L14:2173 suppresses
  18. L12:2097 excites
  19. L11:1334 excites
  20. L11:2278 suppresses
  21. L11:1284 suppresses
  22. L15:2766 excites
  23. L10:1126 excites
  24. L11:44 excites

Feeds later layers

  1. L20:2643 strongly excites
  2. L18:1898 excites
  3. L18:3049 excites
  4. L19:2460 excites
  5. L18:2353 excites
  6. L19:3069 excites
  7. L18:3001 excites
  8. L17:1449 suppresses
  9. L19:2906 excites
  10. L20:1054 suppresses
  11. L19:2825 excites
  12. L18:1768 suppresses
  13. L17:3005 excites
  14. L17:1162 excites
  15. L19:2853 excites
  16. L17:1621 excites
  17. L18:2777 suppresses
  18. L19:1626 suppresses
  19. L21:1615 suppresses
  20. L22:181 excites
  21. L17:1986 excites
  22. L17:1004 suppresses
  23. L25:1473 suppresses
  24. L19:1795 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 -4.18 and -3.921 values between -3.66 and -3.403 values between -3.40 and -3.151 values between -2.89 and -2.633 values between -2.63 and -2.371 values between -2.37 and -2.114 values between -2.11 and -1.857 values between -1.85 and -1.606 values between -1.60 and -1.3426 values between -1.34 and -1.0832 values between -1.08 and -0.8275 values between -0.82 and -0.56268 values between -0.56 and -0.30878 values between -0.30 and -0.041734 values between -0.04 and 0.22423 values between 0.22 and 0.47204 values between 0.47 and 0.73103 values between 0.73 and 0.9952 values between 0.99 and 1.2538 values between 1.25 and 1.5118 values between 1.51 and 1.7712 values between 1.77 and 2.029 values between 2.02 and 2.281 values between 2.28 and 2.543 values between 2.54 and 2.801 values between 2.80 and 3.061 values between 3.06 and 3.322 values between 4.87 and 5.131 values between 5.90 and 6.16offnegativepositiveoften

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. 被判
  2. eyer
  3. graded
  4. 提供的
  5. 被
  6. reak
  7. PCS
  8. 来源
  9. ificant
  10. pora

Suppresses

  1. strategy
  2. approach
  3. attempt
  4. strategies
  5. →defer
  6. anoia
  7. effort
  8. Strategy
  9. endeavor
  10. Strategy