Neuron Explorer

Qwen3-0.6B

'think' after 'I' or 'I don't'

A neuron in layer 5 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 word 'think' in first-person speech, such as 'I think' or 'I don't think', usually in quoted remarks. Turning it up gave unrelated text.

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

    if you serve one to them on a plate."⏎Shaunfield said what kids learn in the garden they're likely to bring back home to their parents.⏎"I think

  2. #2 peak

    xic patients that anything is wrong with them, Dwyer told LiveScience.⏎“Unless we conceive of it as more of a metabolic function, I don’t think we’ll get past the

  3. #3 peak

    The studio probably had little choice but to create a form letter.⏎I don’t think the form letter nature of the rejection is necessarily impersonal. It sticks to objective

  4. #4 peak

    in our children:⏎|Image: San José Library on Flickr|⏎2. DRAW! My favorite gift to give at birthday parties is art supplies. I don’t think kids can every have enough drawing paper, pencils, crayons, markers, or

  5. #5 peak

    “In fact you can’t even differentiate, you can’t even separate the tech from the social. They are irretrievably interwoven together. And I think this is the biggest problem with these kinds of projects, and with technology more generally

  6. #6 peak

    Read more about some trail-blazing female dentists, as well as the forward-thinking men who supported women in the profession:⏎Breaking the Barrier⏎I don’t think many of my professional brethren like it much that

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: a cake and a cake of a cake. Let's say the number of the cake is 3, and the number of the other cake. Let's say the number of the other cake. Let's say the number of the other cake

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

    been developed over many years, thus demonstrating an influence of past and future cases. Similarly, the doctrine of precedent ensures that similar cases have similar results. Procedural grounds for objections protect witnesses from harassment and potential confusion. Therefore, all of these

  2. #2 peak

    can cause agitation, anxiety, insomnia, increased heart rate and blood pressure, and even psychosis.⏎Adderall is a highly-addictive amphetamine whose potential side effects include hallucinations,

  3. #3 peak

    Jim Sublett, chief of allergy immunology at the University of Louisville School of Medicine in Kentucky has repeatedly said in literature that over-the-counter remedies can have adverse side effects.⏎The side effects

  4. #4 peak

    - Faster passenger boarding at the PTI can aid station operators in running a safer and quicker service.⏎- (Side effect) improved passenger movement will help to reduce larger crowds and subsequently improving safety.<|endoftext|>

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. L4:1694 weakly excites
  2. L3:1866 weakly excites
  3. L0:3006 weakly suppresses
  4. L1:3057 weakly suppresses
  5. L4:2695 weakly excites
  6. L2:2050 weakly excites
  7. L4:2188 weakly excites
  8. L1:1027 weakly suppresses
  9. L2:2522 weakly excites
  10. L3:2624 weakly excites
  11. L0:1006 weakly suppresses
  12. L1:1504 weakly suppresses
  13. L4:2193 weakly suppresses
  14. L1:2812 weakly excites
  15. L0:531 weakly excites
  16. L4:730 weakly excites
  17. L0:234 weakly excites
  18. L2:1598 weakly suppresses
  19. L3:1460 weakly excites
  20. L2:864 weakly excites
  21. L2:2101 weakly excites
  22. L3:1048 weakly excites
  23. L4:2515 weakly excites
  24. L3:540 weakly excites

Feeds later layers

  1. L6:2740 suppresses
  2. L9:401 weakly suppresses
  3. L12:1732 weakly suppresses
  4. L10:1324 weakly suppresses
  5. L13:2825 weakly excites
  6. L19:1545 weakly suppresses
  7. L13:1883 weakly suppresses
  8. L8:660 weakly excites
  9. L25:1274 weakly excites
  10. L10:2094 weakly suppresses
  11. L13:210 weakly suppresses
  12. L17:71 weakly excites
  13. L12:1854 weakly suppresses
  14. L17:68 weakly suppresses
  15. L9:917 weakly suppresses
  16. L8:423 weakly suppresses
  17. L6:785 weakly suppresses
  18. L13:899 weakly suppresses
  19. L11:300 weakly suppresses
  20. L7:2248 weakly excites
  21. L14:562 weakly suppresses
  22. L13:738 weakly suppresses
  23. L27:1818 weakly suppresses
  24. L11:2158 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 -0.67 and -0.632 values between -0.58 and -0.543 values between -0.54 and -0.501 values between -0.50 and -0.466 values between -0.46 and -0.4111 values between -0.41 and -0.3721 values between -0.37 and -0.3335 values between -0.33 and -0.2956 values between -0.29 and -0.24111 values between -0.24 and -0.20142 values between -0.20 and -0.16249 values between -0.16 and -0.12371 values between -0.12 and -0.07519 values between -0.07 and -0.03752 values between -0.03 and 0.01568 values between 0.01 and 0.06410 values between 0.06 and 0.10265 values between 0.10 and 0.14182 values between 0.14 and 0.1887 values between 0.18 and 0.2356 values between 0.23 and 0.2723 values between 0.27 and 0.3120 values between 0.31 and 0.3510 values between 0.35 and 0.402 values between 0.40 and 0.441 values between 0.44 and 0.481 values between 0.48 and 0.521 values between 0.57 and 0.611 values between 0.61 and 0.651 values between 0.99 and 1.04offnegativepositiveoften

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. rall
  2. ooke
  3. leitung
  4. erox
  5. ière
  6. `\
  7. elman
  8. .weixin
  9. uiltin
  10. scient

Suppresses

  1. Her
  2. еств
  3. /her
  4. idential
  5. her
  6. 都市报
  7. 叆
  8. unks
  9. 垓
  10. her