Neuron Explorer

Qwen3-0.6B

'by' or operator after a number in arithmetic

A neuron in layer 11 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 tends to fire on the word just before a maths step, especially 'by' in instructions like 'multiply the number of stairs by' or 'divide the total weight by'. When turned up, one answer slipped into multiplying numbers, which fits, but most of the text was unrelated.

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

    a comment if you LOVE your math series! I'd love to check it out!⏎I like the tricks (9s trick, Doubles, Doubles Plus 1) that Saxon uses to teach facts, but even after learning and practicing

  2. #2 peak

    - Determine the height of the landing from the bottom of the stairs. this will⏎- Decide the number of stairs and their width. multiply the number of stairs by .28 to get the total landing space required⏎1. Cascading deck

  3. #3 peak

    TMJ disorder develops, signs and symptoms to watch out for, and tmj treatment Albuquerque options to address the condition effectively.⏎What is TMJ?⏎Temporomandibular joint disorder, or TMJD, is a condition that affects a

  4. #4 peak

    s–1 ( Eddington limit for neutron stars, ), and then typically decaying in a few tens of seconds3.⏎Multiplying the burst luminosity by its duration we get an estimate of the total burst energy 1039

  5. #5 peak

    illin . Tell your doctor or pharmacist if you have any medical conditions, especially if any of the following apply to you:⏎if you are pregnant, planning to become pregnant, or are breast-feeding⏎if you are taking any prescription or nonpres

  6. #6 peak

    .⏎- Add an additional 25 percent of the total weight to allow a margin of error.⏎- Divide the total weight including the additional 25 percent by the number of wheels supporting the cart. The resulting

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 most common, most accurate, and even, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, 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

    fairly vanilla neural networks called Alice, Bob, and Eve.⏎Ref: Learning to Protect Communications with Adversarial Neural Cryptography. arXiv - Computer Science > Cryptography and Security (21 October 2016) | ar

  2. #2 peak

    leader for fourteen years.⏎Subjects: History — Social Sciences.<|endoftext|>Newegg.com - A great place to buy computers, computer

  3. #3 peak

    , his own understanding of them.⏎Subjects: Judaism and Jewish Studies.<|endoftext|>Did you ever wonder the meaning of the word Christmas

  4. #4 peak

    kworm-BEVS is a promising alternative strategy to produce the functional and thermostable reverse transcriptase.⏎All Science Journal Classification (ASJC) codes⏎- Insect Science<|endoftext|>An account of the religion of

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. L4:303 suppresses
  2. L9:679 excites
  3. L3:164 weakly suppresses
  4. L7:1756 weakly suppresses
  5. L8:746 weakly suppresses
  6. L6:1674 weakly excites
  7. L9:1206 weakly suppresses
  8. L10:1110 weakly suppresses
  9. L10:891 weakly suppresses
  10. L9:327 weakly excites
  11. L3:348 weakly suppresses
  12. L2:862 weakly excites
  13. L6:1610 weakly excites
  14. L8:881 weakly suppresses
  15. L6:158 weakly excites
  16. L8:1083 weakly excites "x" or "by" between two dimensions
  17. L4:2574 weakly excites
  18. L1:2926 weakly suppresses
  19. L10:1285 weakly suppresses
  20. L4:378 weakly excites
  21. L10:1331 weakly suppresses
  22. L5:2771 weakly excites
  23. L10:2607 weakly suppresses
  24. L9:2334 weakly excites

Feeds later layers

  1. L14:1654 weakly excites
  2. L16:2196 weakly excites
  3. L19:2698 weakly excites
  4. L17:2322 weakly excites
  5. L18:1637 weakly excites
  6. L16:1818 weakly excites "two" and the link between two things
  7. L14:1082 weakly suppresses
  8. L16:490 weakly excites
  9. L17:3010 weakly excites
  10. L23:2547 weakly excites
  11. L20:2005 weakly suppresses
  12. L17:1321 weakly excites
  13. L17:1036 weakly excites
  14. L18:1887 weakly excites
  15. L15:1078 weakly suppresses
  16. L20:1001 weakly excites
  17. L16:2323 weakly suppresses
  18. L16:2856 weakly suppresses
  19. L12:944 weakly suppresses
  20. L19:2323 weakly suppresses
  21. L17:462 weakly suppresses
  22. L12:1978 weakly suppresses
  23. L18:478 weakly suppresses
  24. L18:685 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 -3.25 and -3.091 values between -2.13 and -1.971 values between -1.81 and -1.651 values between -1.65 and -1.502 values between -1.50 and -1.341 values between -1.34 and -1.183 values between -1.18 and -1.027 values between -1.02 and -0.866 values between -0.86 and -0.7025 values between -0.70 and -0.5438 values between -0.54 and -0.38151 values between -0.38 and -0.22679 values between -0.22 and -0.062484 values between -0.06 and 0.10436 values between 0.10 and 0.2654 values between 0.26 and 0.429 values between 0.42 and 0.584 values between 0.58 and 0.741 values between 0.74 and 0.901 values between 0.90 and 1.061 values between 1.21 and 1.371 values between 1.85 and 2.011 values between 2.97 and 3.13offnegativepositiveoften

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. `
  2. omb
  3. ~⏎⏎
  4. 印
  5. (:,:,
  6. `(
  7. perceive
  8. Providence
  9. onium
  10. absorb

Suppresses

  1. olarity
  2. ẩm
  3. marks
  4. rier
  5. ực
  6. ihan
  7. ramento
  8. athing
  9. asher
  10. lean