Neuron Explorer

Qwen3-0.6B

unclear: sentence ends in factual prose

A neuron in layer 11 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 only weakly, mostly on full stops and small joining words, such as the end of a factual sentence or "so that". The pattern is not clear enough to name a topic, and the steering did not show a clear theme.

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

    of high mortality, this trend is expected to accelerate. Around 2030, this could lead to a situation where there would be more deaths than births. If fertility increases to 1.7 children per woman, the point when the

  2. #2 peak

    the base if cut back by fire[⏎The nectar is used to make a sweet drink[⏎The roots and bark are antipyretic and antiseptic. They are used traditionally to treat headaches[⏎If you have any useful information about this

  3. #3 peak

    Finally, make sure that your goal has a timeline attached so that you can measure progress over time and stay motivated throughout the process of achieving it.⏎

  4. #4 peak

    . People at risk for kidney stones may increase that risk; people may also have diarrhea.⏎The Institute of Medicine has determined the maximum limits for 24 nutrients. This table is for adults 19 years of age or older. It does

  5. #5 peak

    and promotes regular bowel movements. Eating a 100-gram serving of tamales introduces 6.28 grams of protein into your diet. This is 13 percent of the daily value for protein, a nutrient your body

  6. #6 peak

    Google.⏎“We think we will have fourth-generation fuels in about 18 months, with CO2 as the fuel stock. We have 20 million genes which I call the design components of the future,” Venter said. “We are

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: and I think of it as the most of it. to the world of the world. It's a thing of it. But the world of the world, to the world of the world, to the world of the world, to 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

    Miami Physics Department maintains an⏎Observatory for teaching purposes. The Copernicus Observatory is located⏎on the Coral Gables Campus, on the roof of the⏎Ungar Computing<|endoftext|>Higher insulation R-values, higher U-value windows, and

  2. #2 peak

    uffed at the knees, a style that⏎involves folding the garment in half and tucking it under the waist. This style⏎of draping the⏎dhoti is known as the half-fold. The dhoti can be

  3. #3 peak

    get reactivated any time. When the HSV-1 is reactivated, cold sores appear. Some of the more common triggers for its reactivation and, hence, the formation of cold sores, include the following.⏎- Psychological and

  4. #4 peak

    hungry, we are more likely to overeat or choose unhealthy options. Reduce the chance of this happening with the following tips:⏎- Bring your coffee from home; to avoid temptation in cafés or shops e.g. croissants, muffins

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. L10:816 excites
  2. L10:2104 excites opening parenthesis in formulas and citations
  3. L7:288 excites
  4. L6:244 excites
  5. L9:819 suppresses
  6. L6:710 excites
  7. L9:1659 excites
  8. L8:88 excites line breaks inside wrapped sentences
  9. L8:692 suppresses
  10. L9:950 excites
  11. L9:1055 suppresses
  12. L10:2993 excites
  13. L10:612 excites full stop ending boilerplate or a title
  14. L9:1942 suppresses
  15. L6:1126 suppresses
  16. L5:828 excites
  17. L6:912 excites
  18. L7:2752 suppresses
  19. L6:179 excites
  20. L6:854 excites comma after an introductory word like then or when
  21. L4:312 weakly suppresses 'that' after 'the way' or 'the first time'
  22. L9:268 weakly suppresses
  23. L8:1656 weakly suppresses
  24. L8:2208 weakly suppresses

Feeds later layers

  1. L14:1274 excites
  2. L16:995 excites
  3. L13:2444 excites
  4. L14:1318 excites "both" and "either" before alternatives
  5. L15:885 excites
  6. L13:2903 excites
  7. L13:1578 excites
  8. L16:2158 excites
  9. L12:2811 weakly excites
  10. L15:1794 weakly excites
  11. L14:155 weakly excites sentence-ending periods in prose
  12. L12:1848 weakly suppresses
  13. L18:2811 weakly excites
  14. L17:2323 weakly excites
  15. L16:718 weakly excites
  16. L18:2202 weakly excites
  17. L13:1734 weakly excites sentence end right before a line break
  18. L24:1273 weakly excites
  19. L20:763 weakly excites
  20. L13:1698 weakly excites
  21. L24:1347 weakly suppresses
  22. L12:1585 weakly excites
  23. L17:1204 weakly excites
  24. L26:2602 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.

1 values between -2.28 and -2.192 values between -2.10 and -2.021 values between -1.67 and -1.592 values between -1.59 and -1.501 values between -1.50 and -1.421 values between -1.42 and -1.332 values between -1.33 and -1.241 values between -1.24 and -1.162 values between -1.16 and -1.072 values between -0.99 and -0.901 values between -0.90 and -0.816 values between -0.81 and -0.735 values between -0.73 and -0.644 values between -0.64 and -0.569 values between -0.56 and -0.4718 values between -0.47 and -0.3862 values between -0.38 and -0.30209 values between -0.30 and -0.21473 values between -0.21 and -0.13921 values between -0.13 and -0.041367 values between -0.04 and 0.05514 values between 0.05 and 0.13179 values between 0.13 and 0.2263 values between 0.22 and 0.3030 values between 0.30 and 0.3910 values between 0.39 and 0.487 values between 0.48 and 0.564 values between 0.56 and 0.654 values between 0.65 and 0.732 values between 0.73 and 0.822 values between 0.82 and 0.912 values between 0.91 and 0.991 values between 1.08 and 1.16offnegativepositiveoften

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. ikal
  2. Messiah
  3. лага
  4. naï
  5. stä
  6. asive
  7. KERNEL
  8. nah
  9. oload
  10. 鏡

Suppresses

  1. .appspot
  2. ubit
  3. �单
  4. ARC
  5. isSelected
  6. billions
  7. ighthouse
  8. TRL
  9. OCR
  10. werde