Neuron Explorer

Qwen3-0.6B

full stops ending informational sentences

A neuron in layer 16 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 full stop at the end of a plain informational sentence, especially one that introduces a point that the next sentence will explain, such as 'Cloud providers often have multiple data centres.' It is slightly negative on punctuation that cuts a sentence short. Turning it up gave repetitive or off-topic text rather than clear explanatory writing.

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

    measures safeguard against physical threats like theft, fire, and natural disasters.⏎Redundancy and Backup⏎Cloud providers often have multiple data centres located in different geographical locations. This redundancy ensures that your data is backed up and accessible even if one data centre

  2. #2 peak

    predictive accuracy of the model.⏎A. Basic Structure of Decision Trees⏎The basic structure of a decision tree consists of a root node, branches, and leaf nodes. The root node represents the top of the tree, and it contains all the instances

  3. #3 peak

    skills but also foster a love for scientific inquiry.⏎Cultivate Creativity with Arts and Crafts⏎Artistic expression plays a vital role in a child’s development. Encourage your students

  4. #4 peak

    infrastructure investments. This flexibility enables you to maintain a secure environment as your business expands.⏎Compliance and Regulation⏎Cloud providers adhere to industry-specific compliance standards and regulations. They implement security measures that

  5. #5 peak

    works toward 2025 industry goals⏎Every five years, Field to Market: The Alliance for Sustainable Agriculture publishes its National Indicators Report (NIR). The latest

  6. #6 peak

    key health benefits of fenugreek leaves is their high fiber content. Dietary fiber is an important nutrient that helps to promote healthy digestion and prevent constipation

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 problem of. In some cases, we have a lot of problems that are difficult to solve. I think that most people would have a lot of problems. I also have some of the problems that I have. So I need to have

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

    is the main cause of global warming. So it must be recycled for a healthy environment. I try to give some ways to recycle the butts which are as follow:⏎1. Bed sheet and pillow with cigarette butts⏎This is an innovative

  2. #2 peak

    .⏎We currently use Saxon Math. Next year, we will start looking at different math series so PLEASE, PLEASE leave a comment if you LOVE your math series! I'd love to check it out!⏎I like the tricks (9s trick

  3. #3 peak

    volume as the cube. What is your hypothesis? Will the ice cube or crushed ice melt faster, or will they melt at the same rate? Explain your reasoning.⏎Poll the students on their hypotheses. Record and display their responses for all to see

  4. #4 peak

    steel plate is called galvanized steel composite panel. What are the manufacturing processes of galvanized steel composite panel? How to choose a good galvanized steel composite panel?⏎1. Hot-dip galvanized steel sheet. Dip the steel sheet into a

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. L9:16 suppresses unclear: periods inside numbers and names
  2. L9:12 suppresses
  3. L15:975 suppresses
  4. L14:155 excites sentence-ending periods in prose
  5. L9:2447 excites
  6. L11:132 excites period ending an opening topic sentence
  7. L4:1546 suppresses
  8. L14:634 excites
  9. L8:304 suppresses
  10. L1:1 excites
  11. L0:80 suppresses
  12. L12:2141 suppresses
  13. L13:3004 suppresses
  14. L8:32 excites
  15. L8:361 excites
  16. L14:1128 excites
  17. L7:259 suppresses
  18. L14:1142 excites
  19. L2:156 weakly excites
  20. L5:3 weakly suppresses
  21. L8:1515 weakly excites
  22. L10:478 weakly suppresses
  23. L13:2614 weakly excites
  24. L0:567 weakly suppresses

Feeds later layers

  1. L18:2837 weakly excites
  2. L20:2551 weakly excites
  3. L19:3040 weakly excites
  4. L24:2885 weakly excites
  5. L25:1503 weakly excites
  6. L24:2610 weakly excites
  7. L17:88 weakly suppresses
  8. L18:1370 weakly excites
  9. L19:1411 weakly excites
  10. L21:1947 weakly excites
  11. L20:1251 weakly excites
  12. L23:211 weakly excites
  13. L18:2956 weakly suppresses
  14. L17:2075 weakly excites
  15. L23:2036 weakly excites
  16. L27:1193 weakly excites
  17. L26:1981 weakly excites
  18. L18:984 weakly excites
  19. L17:612 weakly excites
  20. L24:2726 weakly excites
  21. L26:1888 weakly excites
  22. L19:2919 weakly excites
  23. L23:1812 weakly suppresses
  24. L17:68 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 -4.57 and -4.292 values between -3.74 and -3.462 values between -3.46 and -3.183 values between -3.18 and -2.902 values between -2.90 and -2.628 values between -2.62 and -2.346 values between -2.34 and -2.077 values between -2.07 and -1.799 values between -1.79 and -1.5114 values between -1.51 and -1.2329 values between -1.23 and -0.9536 values between -0.95 and -0.67107 values between -0.67 and -0.39628 values between -0.39 and -0.122461 values between -0.12 and 0.16431 values between 0.16 and 0.4459 values between 0.44 and 0.7226 values between 0.72 and 1.0019 values between 1.00 and 1.2811 values between 1.28 and 1.567 values between 1.56 and 1.835 values between 1.83 and 2.1112 values between 2.11 and 2.396 values between 2.39 and 2.675 values between 2.67 and 2.953 values between 2.95 and 3.234 values between 3.23 and 3.501 values between 3.50 and 3.782 values between 4.34 and 4.621 values between 6.29 and 6.57offnegativepositiveoften

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. umes
  2. ucion
  3. IonicModule
  4. ');?></
  5. Laden
  6. itudes
  7. ')?></
  8. setattr
  9. issors
  10. ObjectContext

Suppresses

  1. :
  2. 击
  3. Ab
  4. lias
  5. 大量
  6. An
  7. .has
  8. “
  9. .An
  10. needs