Neuron Explorer

Qwen3-0.6B

decimal point or digit inside a number

A neuron in layer 15 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 reacts to the dot in numbered sections like '3.2' or '5.9 g', and it is pushed the other way by digits inside numbers, such as the 0 in '10'. The effect is weak, and turning it up or down made the writing change in unrelated ways.

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

    two end systems⏎B. It specifies the quality of the data & connection⏎C. It specify the size of data⏎D. All of the above⏎6-4. The class-based addressing is also known

  2. #2 peak

    3.2), the sash should be positioned to produce the recommended face velocity, which

  3. #3 peak

    3. And the map of the world, which colours a country blue if it's leaning

  4. #4 peak

    3. Demonstration on pest management technologies like pheromone traps, fruit bagging

  5. #5 peak

    3.5 million African-American slaves, and now the Union Army began to recruit a second

  6. #6 peak

    3.16. The consolidation and merging of data sources provides multiple advantages for compilers.

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 been working on the problem of the time period in a math class, but I'm stuck, but I've been stuck for a long time. I'm thinking of, and in the hour of the class, but I'm stuck

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

    different, which makes remembering them a difficult task. The following are some creative ways that can help you create new passwords and actually remember them all.⏎Remember 100 different passwords with 1 rule⏎One way to generate unique passwords that you

  2. #2 peak

    /10.5479/si.GVP.BGVN200701-333060⏎28.98°S, 140.25°W; summit elev

  3. #3 peak

    mm;⏎9. External material⏎1.5mm thick cold-rolled steel plate with paint treatment;⏎10. Box door

  4. #4 peak

    definition for<|endoftext|>I have recently been re-reading two books I purchased a few years ago, when I was still teaching languages traditionally. The books are:⏎100+ Fun Ideas for Practising Modern Foreign Languages in the Primary Classroom and⏎

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. L14:1 excites
  2. L14:375 suppresses
  3. L14:7 suppresses
  4. L14:402 excites
  5. L14:9 suppresses
  6. L4:647 weakly excites
  7. L14:817 weakly excites
  8. L3:485 weakly excites
  9. L7:576 weakly excites
  10. L14:784 weakly suppresses
  11. L7:1060 weakly excites
  12. L14:33 weakly suppresses
  13. L13:595 weakly suppresses
  14. L13:695 weakly excites
  15. L3:2542 weakly suppresses
  16. L7:310 weakly excites
  17. L13:933 weakly suppresses
  18. L7:728 weakly excites
  19. L14:921 weakly excites
  20. L14:338 weakly suppresses
  21. L13:925 weakly suppresses
  22. L13:534 weakly suppresses unclear: sentence-ending periods
  23. L10:636 weakly excites unclear: 'your' in contact boilerplate
  24. L9:630 weakly excites

Feeds later layers

  1. L16:2152 suppresses
  2. L16:1695 weakly suppresses
  3. L16:1258 weakly excites
  4. L20:685 weakly excites
  5. L17:2406 weakly excites
  6. L19:1839 weakly excites
  7. L18:707 weakly excites
  8. L17:2402 weakly suppresses
  9. L27:1341 weakly suppresses
  10. L18:660 weakly suppresses
  11. L16:521 weakly suppresses
  12. L17:570 weakly suppresses
  13. L19:1792 weakly suppresses word just before a 'reason is' explanation
  14. L20:1917 weakly excites
  15. L22:443 weakly suppresses
  16. L20:1036 weakly excites
  17. L17:295 weakly suppresses
  18. L20:2950 weakly excites
  19. L17:1155 weakly excites opening parenthesis after a sentence
  20. L18:2162 weakly excites
  21. L23:2205 weakly suppresses
  22. L16:2440 weakly suppresses
  23. L25:1675 weakly excites
  24. L19:2562 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 -1.36 and -1.311 values between -1.31 and -1.261 values between -1.10 and -1.042 values between -1.04 and -0.992 values between -0.99 and -0.941 values between -0.94 and -0.892 values between -0.89 and -0.835 values between -0.83 and -0.782 values between -0.78 and -0.734 values between -0.73 and -0.688 values between -0.68 and -0.6216 values between -0.62 and -0.5714 values between -0.57 and -0.5242 values between -0.52 and -0.4739 values between -0.47 and -0.4164 values between -0.41 and -0.3690 values between -0.36 and -0.31141 values between -0.31 and -0.25206 values between -0.25 and -0.20256 values between -0.20 and -0.15308 values between -0.15 and -0.10361 values between -0.10 and -0.04558 values between -0.04 and 0.01450 values between 0.01 and 0.06374 values between 0.06 and 0.11288 values between 0.11 and 0.17203 values between 0.17 and 0.22164 values between 0.22 and 0.27111 values between 0.27 and 0.3389 values between 0.33 and 0.3843 values between 0.38 and 0.4325 values between 0.43 and 0.4821 values between 0.48 and 0.5411 values between 0.54 and 0.593 values between 0.59 and 0.641 values between 0.64 and 0.691 values between 0.69 and 0.75offnegativepositiveoften

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. aps
  2. f
  3. igos
  4. 改进
  5. F
  6. SEP
  7. apa
  8. (F
  9. expanded
  10. Sept

Suppresses

  1. 维奇
  2. ��
  3. 记者在
  4. liğinde
  5. 苟
  6. configFile
  7. nevertheless
  8. 😉⏎⏎
  9. linger
  10. pressions