Neuron Explorer

Qwen3-0.6B

unclear: measured quantities in technical prose

A neuron in layer 14 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 words in scientific or statistical writing that name a measured thing, such as 'age', 'concentration', 'energy' or 'duration'. It seems to prefer words that lead into a quantity. Turning it up gave unrelated text.

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

    or home fire creature.⏎Already this difference between ancient settlements and younger settlements (settlements that are only one thousand years old!), gives us a hint about the age of the Lokke-notions. Besides from that, it is quite possible that

  2. #2 peak

    , compared with control groups. The studies were all short term — mostly lasting two to eight weeks — so it’s not clear how long the blood pressure benefits would last, but the “small reduction we saw over the short term might complement other treatment

  3. #3 peak

    “While we don’t yet know why THC influences the female reproductive system, we do know that the response appears to be dose dependent. This means that the higher concentration of THC that is being used, the more the reproductive system is affected,” she

  4. #4 peak

    below the official poverty line (OPL). The figure was 15.2 per cent in 2006. This shows a commendable reduction of poverty. The Department of Census and Statistics (DCS) declares that the O

  5. #5 peak

    (the standard approach for studying the microbiome). Sampling was done over time, so the temporal dynamics of the disease could also be visualized. Like previous studies, the group

  6. #6 peak

    slowly than simple sugars in white bread, for example. The slower digestion means that all of the carbohydrates aren’t released into the bloodstream at once and provide a more prolonged source of fuel.⏎-During races, however, the goal is to replenish

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 first few days of the week. The first few days of the week are the first few days. The first few days are the first few days of the week. The first day of the week are the first few days of the week

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

    30. It tells you the oil’s thickness or viscosity. A thinner motor oil — indicated by a lower number — flows more easily. Conversely, a thicker oil — with a larger number —

  2. #2 peak

    manner based on space indentation. Place major points farthest to the left. Indent each more specific point farther to the right (level of importance is indicated by distance away from left margin).⏎What is outline format?⏎An outline is a tool used

  3. #3 peak

    specificity of 25% to 80%. Typically a fracture can be seen on ultrasonography as a disruption in the hyperechoic linear cortex of bone as well as surrounding hemat

  4. #4 peak

    by glass it is either shifted away by air movement or re-radiated by the glass surface.⏎Typically, a highly reflective material will have a low emissivity, and a dull darker colored material will have a high emissivity. Reducing

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. L12:2549 weakly suppresses
  2. L13:1407 weakly suppresses
  3. L7:1293 weakly excites
  4. L13:2232 weakly excites
  5. L12:2131 weakly excites
  6. L11:110 weakly suppresses
  7. L13:881 weakly suppresses
  8. L4:239 weakly excites
  9. L11:1421 weakly suppresses
  10. L0:118 weakly suppresses
  11. L13:1242 weakly excites
  12. L7:1646 weakly suppresses
  13. L7:162 weakly suppresses
  14. L5:258 weakly excites
  15. L12:285 weakly excites
  16. L13:2318 weakly excites
  17. L7:2407 weakly suppresses
  18. L13:3037 weakly suppresses
  19. L11:1248 weakly excites
  20. L7:2005 weakly excites
  21. L13:86 weakly suppresses
  22. L11:1494 weakly suppresses
  23. L8:160 weakly excites
  24. L2:1008 weakly excites

Feeds later layers

  1. L15:1943 suppresses
  2. L16:1474 weakly excites
  3. L15:163 weakly suppresses
  4. L15:185 weakly excites
  5. L16:2734 weakly suppresses
  6. L15:1714 weakly excites
  7. L18:2249 weakly excites
  8. L16:1129 weakly excites
  9. L17:33 weakly suppresses
  10. L15:2037 weakly excites
  11. L16:1229 weakly excites
  12. L27:741 weakly excites
  13. L17:1328 weakly excites
  14. L18:433 weakly excites
  15. L22:564 weakly excites
  16. L16:2229 weakly excites
  17. L27:796 weakly excites
  18. L16:2018 weakly suppresses
  19. L15:515 weakly excites word before 'that', 'to' or 'for' continuation
  20. L17:2716 weakly excites
  21. L15:1209 weakly suppresses
  22. L18:1621 weakly excites
  23. L15:409 weakly excites
  24. L22:1678 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 -2.91 and -2.781 values between -2.16 and -2.031 values between -1.65 and -1.531 values between -1.40 and -1.282 values between -1.28 and -1.152 values between -1.15 and -1.022 values between -1.02 and -0.904 values between -0.90 and -0.775 values between -0.77 and -0.6516 values between -0.65 and -0.5219 values between -0.52 and -0.4042 values between -0.40 and -0.27172 values between -0.27 and -0.14943 values between -0.14 and -0.022135 values between -0.02 and 0.11429 values between 0.11 and 0.2395 values between 0.23 and 0.3615 values between 0.36 and 0.4913 values between 0.49 and 0.615 values between 0.61 and 0.742 values between 0.74 and 0.862 values between 0.86 and 0.991 values between 1.99 and 2.12offnegativepositiveoften

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. ifax
  2. 却被
  3. itics
  4. Newcastle
  5. 險
  6. לו
  7. -risk
  8. 险
  9. iu
  10. <!--[

Suppresses

  1. мен
  2. semb
  3. 矛
  4. enty
  5. OCI
  6. 茑
  7. sass
  8. ��
  9. 猗
  10. _ABS