Neuron Explorer

Qwen3-0.6B

unclear: mixed mid-sentence words

A neuron in layer 15 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 a grab bag of words in the middle of sentences, with no clear shared theme. When it was turned up, the model drifted into unrelated topics like math problems and technical documents.

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

    smaller rectangles. How many different ways can you fold⏎This problem focuses on Dienes' Logiblocs. What is the same and⏎what is different about these pairs of shapes? Can you describe the⏎shapes in the picture?⏎10

  2. #2 peak

    Once this has been learned, small bundles of neurons in the spinal cord - the central pattern generators (CPG) - ensure that this sequence happens almost automatically: we do not need to think about when and how far away we should place our foot

  3. #3 peak

    element, most often a form of uranium, is split into smaller pieces, releasing energy. As each fuel atom splits, it releases more neutrons, which split yet more atoms. This is a nuclear chain reaction. Water both keeps the reactor cool

  4. #4 peak

    4 (Thyroxine)⏎What it is: T4 is the main hormone produce by the thyroid gland. Free T4 is the amount of available or ‘active’ T4 in the body. Although there is more T4 in the

  5. #5 peak

    as hard as you can. Now, place three fingers from your left hand in the palm of your right hand and squeeze. Notice how much easier it is to squeeze by adding

  6. #6 peak

    the combustible material until ignition occurs. Generally, the build up of heat to ignition point occurs when the material is in a pile so that the heat being generated cannot adequately escape.⏎For ignition to occur, the rate of air movement must be sufficient

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: if I were to put something in my hand, I would like it to be something I might feel, as if I were walking in the same path as the one I might walk in the one I might do, like if I were to

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

    off from their mouth.⏎Jeremiah 22:21 "I spoke to you in your<|endoftext|>How’s your facility looking this flu season? According to the U.S

  2. #2 peak

    vent Reflections: Week 2, Day 5⏎“Moses said to his father-in-law, “Because the people come to me to inquire of God. When they have a dispute, it comes to me, and I judge between

  3. #3 peak

    Holman Christian Standard BibleJesus told him, “Foxes have dens, and birds of the sky have nests, but the Son of Man has no place to lay His head.”⏎New American Standard BibleAnd Jesus

  4. #4 peak

    anger.” 13 And suddenly there was with the angel a multitude of the heavenly host praising God and saying,⏎14 “Glory to God in the highest,⏎and on earth peace among those with whom he is pleased!”⏎15

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. L10:1773 suppresses
  2. L12:1968 suppresses
  3. L10:1810 suppresses
  4. L10:2385 excites
  5. L9:937 weakly suppresses
  6. L13:1467 weakly excites
  7. L9:748 weakly suppresses
  8. L11:3001 weakly suppresses
  9. L14:1965 weakly excites
  10. L10:2370 weakly suppresses
  11. L11:546 weakly excites
  12. L0:712 weakly suppresses
  13. L13:954 weakly excites
  14. L9:1397 weakly suppresses
  15. L9:1212 weakly excites
  16. L12:2083 weakly suppresses
  17. L14:15 weakly excites
  18. L10:1302 weakly excites
  19. L1:1578 weakly excites
  20. L14:1149 weakly excites
  21. L9:832 weakly suppresses
  22. L13:1466 weakly excites
  23. L1:2339 weakly excites
  24. L7:1379 weakly suppresses

Feeds later layers

  1. L19:1469 suppresses
  2. L24:437 suppresses
  3. L17:1963 suppresses
  4. L20:1089 weakly suppresses
  5. L18:1417 weakly excites
  6. L18:2039 weakly suppresses punctuation that introduces an explanation
  7. L25:1942 weakly suppresses
  8. L20:1117 weakly suppresses bible verses, mid-phrase words
  9. L19:1484 weakly suppresses
  10. L17:1487 weakly suppresses
  11. L24:2494 weakly suppresses
  12. L17:1601 weakly suppresses
  13. L25:502 weakly suppresses
  14. L17:2191 weakly excites
  15. L16:2341 weakly suppresses
  16. L19:1946 weakly suppresses
  17. L19:1401 weakly excites
  18. L16:1720 weakly suppresses
  19. L16:3054 weakly suppresses
  20. L17:2026 weakly suppresses
  21. L20:1359 weakly suppresses
  22. L18:2715 weakly excites
  23. L16:1190 weakly excites
  24. L16:2224 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 -2.04 and -1.942 values between -1.74 and -1.641 values between -1.45 and -1.351 values between -1.35 and -1.252 values between -1.25 and -1.161 values between -1.16 and -1.061 values between -1.06 and -0.965 values between -0.96 and -0.864 values between -0.86 and -0.768 values between -0.76 and -0.6711 values between -0.67 and -0.5714 values between -0.57 and -0.4729 values between -0.47 and -0.3780 values between -0.37 and -0.28223 values between -0.28 and -0.18472 values between -0.18 and -0.081622 values between -0.08 and 0.02844 values between 0.02 and 0.12314 values between 0.12 and 0.21119 values between 0.21 and 0.3162 values between 0.31 and 0.4132 values between 0.41 and 0.5121 values between 0.51 and 0.6113 values between 0.61 and 0.707 values between 0.70 and 0.809 values between 0.80 and 0.903 values between 0.90 and 1.002 values between 1.00 and 1.092 values between 1.09 and 1.191 values between 1.19 and 1.291 values between 1.49 and 1.581 values between 1.78 and 1.88offnegativepositiveoften

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. unresolved
  2. 引用日期
  3. 生产总值
  4. NotSupportedException
  5. ityEngine
  6. Stokes
  7. .SaveChangesAsync
  8. .Support
  9. convolution
  10. matrices

Suppresses

  1. >`;⏎
  2. blick
  3. rua
  4. �
  5. acular
  6. же
  7. "';
  8. �
  9. owers
  10. )}"⏎