Neuron Explorer

Qwen3-0.6B

prepositions right after a noun

A neuron in layer 19 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 small connecting words like 'of', 'on' or 'your' that come straight after a noun or verb and start a phrase. For example, 'the victim of' or 'bored with your'. Turning it up produced repeated filler words, which does not clearly reflect this pattern.

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

    friends and family. They may claim that while on holidays, they have been robbed and all of their money/possessions.⏎The scammer then asks the recipients of the email to send them money, usually via wire or money transfer, so they

  2. #2 peak

    . Knowing which types of exercises fall into which category can help you construct your workouts more effectively, and can give you more options so you don’t get bored with your workouts. When designing your fitness plan, consider adding some stretching and flexibility training.

  3. #3 peak

    , care must be taken in order to avoid or prevent the use of technical jargon. Finally, the language (in this case in reference to the actually language being

  4. #4 peak

    warm, loving and understanding. You show empathy for your child, accept his abilities and are able to take his perspective.⏎What does this mean for the kiddos of authoritative parents? They grow up knowing how to make their own decisions, and are

  5. #5 peak

    death.⏎To assuage this fear of being buried alive the safety coffin was invented. It is fitted with a device, often a bell that allowed the victim of a premature burial to pull a cord that rang the bell placed on the tomb allowing

  6. #6 peak

    Ballotechnics books and related discussion.⏎Suggested Pdf Resources⏎Suggested News Resources⏎Suggested Web Resources⏎Great care has been taken to prepare the information on this page. Elements of the content come from factual and lexical knowledge databases, realm

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 that a certain kind of the real that we are aware of we can transform into the actual that we can transform into its own natural state of its own original form in different different different different different different different different different different different different different

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

    0 year mark.⏎The results found that cancer returned in a quarter of the women given five years of tamoxifen. That fell to 21

  2. #2 peak

    rows of seats in the waiting room. Using the rules, where⏎are they all sitting? Can you find all the possible ways?⏎What is the best way to shunt these carriages so that each train⏎can continue its journey?⏎If

  3. #3 peak

    Wilkes, Robyn Rose, Kathleen Lee, Michael Wilson , John Skinner , and Dewey M. Caron; Bee Informed Partnership)⏎Nearly a Third of Honey Bee Colonies Died in U.S. Last Winter (Agence France-P

  4. #4 peak

    , length and height and apply the relevant formula to calculate and return the volume of a cuboid based on the given dimensions.<|endoftext|>Approximately 35% of the United States population suffers from Diverticulitis.1 Young adults,

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. L18:1120 excites
  2. L17:1009 suppresses
  3. L18:494 excites
  4. L18:638 excites
  5. L18:1298 excites
  6. L15:1800 suppresses
  7. L15:2893 excites
  8. L14:1693 suppresses
  9. L18:1272 excites
  10. L16:1613 suppresses
  11. L16:1643 suppresses
  12. L15:1528 suppresses
  13. L16:2759 excites
  14. L15:2222 suppresses
  15. L14:990 weakly suppresses
  16. L16:2109 weakly suppresses
  17. L17:852 weakly suppresses
  18. L0:437 weakly suppresses
  19. L15:2409 weakly excites
  20. L16:1369 weakly excites
  21. L8:2033 weakly excites
  22. L16:380 weakly excites
  23. L15:2964 weakly suppresses
  24. L17:87 weakly suppresses

Feeds later layers

  1. L20:714 excites
  2. L20:1867 excites
  3. L21:215 suppresses
  4. L20:101 excites
  5. L20:1494 suppresses
  6. L20:254 excites
  7. L20:755 suppresses
  8. L21:989 weakly excites 'the' before a body part or organ
  9. L20:2945 weakly excites
  10. L20:1197 weakly excites 'this' pointing at the current document
  11. L20:1554 weakly suppresses
  12. L20:1012 weakly suppresses
  13. L20:387 weakly excites
  14. L21:696 weakly excites
  15. L20:1983 weakly excites
  16. L24:132 weakly suppresses
  17. L21:2849 weakly excites
  18. L20:2673 weakly suppresses
  19. L21:1669 weakly excites
  20. L24:2963 weakly suppresses
  21. L24:1549 weakly suppresses
  22. L20:1448 weakly suppresses "the" in mid-sentence, weak
  23. L20:1988 weakly excites "the" before a topic noun
  24. L20:364 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.

2 values between -4.82 and -4.491 values between -4.17 and -3.842 values between -3.84 and -3.523 values between -3.52 and -3.194 values between -3.19 and -2.877 values between -2.87 and -2.547 values between -2.54 and -2.2216 values between -2.22 and -1.8926 values between -1.89 and -1.5728 values between -1.57 and -1.2545 values between -1.25 and -0.9288 values between -0.92 and -0.60348 values between -0.60 and -0.271933 values between -0.27 and 0.05895 values between 0.05 and 0.38202 values between 0.38 and 0.7099 values between 0.70 and 1.0358 values between 1.03 and 1.3531 values between 1.35 and 1.6832 values between 1.68 and 2.0020 values between 2.00 and 2.3316 values between 2.33 and 2.6511 values between 2.65 and 2.987 values between 2.98 and 3.308 values between 3.30 and 3.632 values between 3.95 and 4.283 values between 4.28 and 4.603 values between 4.60 and 4.935 values between 4.93 and 5.252 values between 5.25 and 5.572 values between 5.57 and 5.901 values between 7.52 and 7.851 values between 7.85 and 8.17offnegativepositiveoften

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. them
  2. 尤其是
  3. 耶
  4. 売る
  5. output
  6. oup
  7. execution
  8. ลง
  9. R
  10. ('-',

Suppresses

  1. ystone
  2. igator
  3. idential
  4. ensions
  5. arseille
  6. tatus
  7. YPES
  8. ModifiedDate
  9. 呗
  10. okers