Neuron Explorer

Qwen3-0.6B

period ending a sentence that announces a list

A neuron in layer 18 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 sentence that sets up what comes next, such as 'there are also several disadvantages to hard work.' It fits sentences that introduce a list or a set of points. Turned up, the model's text was unrelated and the steering results were mixed.

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

    hard work can often lead to greater mastery of skills, as repetition and practice can lead to improvement over time.⏎However, there are also several disadvantages to hard work. First, it can be time-consuming, leading to burnout and exhaustion. Second

  2. #2 peak

    options.⏎That said, in the past couple of years a number of factors have begun to crystallize in a way that may finally contribute to significant disruption and change. These include falling equipment costs; further innovation in materials;

  3. #3 peak

    sources. Additionally, AI writing assistants can also help improve your writing quality by suggesting improvements to your grammar and formatting.⏎AI has a number of other benefits as well. For example, AI can help you improve your search engine optimization (SEO) efforts

  4. #4 peak

    bite, and most of the time this dies down after a day or two without treatment. However, there are two potentially serious things that need to be ruled out. First, some insect bites can become infected. If this occurs, there may be

  5. #5 peak

    blood may result in bleeding more easily and nose bleeds.⏎More Serious Problems⏎Occasionally, frequent nose bleeds can be a symptom of a more serious problem. They may be an indication of some type of bleeding disorder, which could also cause

  6. #6 peak

    theory begins with cave paintings: In some cave paintings, it’s common to see creatures drawn with far too many limbs.⏎There are a couple of theories behind this. Some suggest that this was simply because the artists had no means of erasing the

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: it's all the more more too, too, too, me too, too, too, Q, too, too,ito, Q, too, h, y,ito, y,ito, y,ito, y, y

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

    <|endoftext|>The liver is an essential organ of our bodies, as it is responsible for producing glucose, breaking down carbohydrates, storing necessary nutrients, and detoxing the body. If a person has any

  2. #2 peak

    were neglected by the government of Assam and the federal authorities: the area had a single college; medical care was poor; infrastructure projects and job opportunities were few. In the nineteen-sixties, the Bodos

  3. #3 peak

    . It recognizes the broad determinants of health – income and poverty, education, environmental factors such as housing and transport – as well as health care and genetic influences. The Society aims to promote the development of scientific knowledge in social medicine. This covers

  4. #4 peak

    .seai.ie<|endoftext|>The peak signal-to-noise ratio (PSNR) is the ratio between a signal's maximum power and the power of the signal's noise. Engineers commonly use the PSNR to measure the quality of reconstructed images that have been

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. L11:132 excites period ending an opening topic sentence
  2. L15:2467 suppresses
  3. L16:1733 suppresses
  4. L6:250 excites
  5. L17:2184 excites
  6. L16:2008 suppresses
  7. L13:263 excites
  8. L15:262 excites
  9. L1:337 weakly suppresses
  10. L10:251 weakly suppresses
  11. L15:945 weakly suppresses
  12. L16:2900 weakly excites
  13. L5:1653 weakly excites
  14. L14:900 weakly suppresses
  15. L13:2103 weakly excites
  16. L17:1490 weakly suppresses
  17. L14:489 weakly suppresses
  18. L16:1361 weakly excites
  19. L13:2937 weakly excites
  20. L14:1761 weakly suppresses
  21. L16:1145 weakly excites commas inside short lists
  22. L14:816 weakly suppresses
  23. L11:513 weakly excites words inside idioms and fixed phrases
  24. L17:1537 weakly suppresses

Feeds later layers

  1. L22:445 strongly suppresses
  2. L27:655 strongly suppresses
  3. L20:2564 suppresses
  4. L22:2150 suppresses
  5. L21:1587 suppresses
  6. L21:1424 suppresses
  7. L19:934 suppresses
  8. L25:637 suppresses
  9. L23:371 suppresses
  10. L20:1684 suppresses
  11. L21:2038 suppresses
  12. L25:2003 suppresses
  13. L19:2110 suppresses
  14. L20:2323 suppresses
  15. L20:2919 suppresses
  16. L26:2774 suppresses
  17. L22:1295 weakly suppresses 'as' before 'well' or 'as much'
  18. L19:3028 weakly suppresses
  19. L21:1973 weakly suppresses
  20. L19:2997 weakly suppresses
  21. L19:2597 weakly suppresses
  22. L20:1806 weakly excites slash inside web addresses
  23. L19:1593 weakly suppresses
  24. L26:124 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.

2 values between -5.23 and -4.801 values between -4.80 and -4.376 values between -4.37 and -3.942 values between -3.94 and -3.513 values between -3.51 and -3.088 values between -3.08 and -2.656 values between -2.65 and -2.2210 values between -2.22 and -1.7916 values between -1.79 and -1.3637 values between -1.36 and -0.9386 values between -0.93 and -0.501151 values between -0.50 and -0.072344 values between -0.07 and 0.36148 values between 0.36 and 0.7942 values between 0.79 and 1.228 values between 1.22 and 1.658 values between 1.65 and 2.085 values between 2.08 and 2.515 values between 2.51 and 2.945 values between 2.94 and 3.376 values between 3.37 and 3.802 values between 3.80 and 4.231 values between 4.23 and 4.662 values between 5.09 and 5.521 values between 5.95 and 6.381 values between 7.24 and 7.671 values between 8.10 and 8.531 values between 11.54 and 11.97offnegativepositiveoften

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. 불구하고
  2. erna
  3. };⏎
  4. }; ⏎
  5. };⏎
  6. нее
  7. }; ⏎
  8. }; ⏎ ⏎
  9. /il
  10. }';⏎

Suppresses

  1. also
  2. also
  3. Also
  4. Also
  5. 也
  6. también
  7. 也要
  8. 也为
  9. также
  10. 同时也