Neuron Explorer

Qwen3-0.6B

'in other words' restatement cues

A neuron in layer 6 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron lights up on the words that introduce a restatement, especially 'words' in 'in other words' and the colon in 'i.e.:'. It is a signal that the next sentence will explain the previous one in simpler terms. Turning it up did not produce a clear theme in the sample text.

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

    ESS is a general topic of conversation everywhere and in every field:- UNITY OF GOD'S ESSENCE, Unity of His attributes, and actions. In other word, also the unity of prophets and their teachings, the unity of the LAW, GH

  2. #2 peak

    that can be hosted by the CLR, while at the same time can be accessed in a uniform manner by all languages that the .NET platform support. In other words, CLS is a subset of the full functionality of CTS.⏎- Common

  3. #3 peak

    or title). If you highlighted a title with the data, you would use this option to prevent the program from sorting your title into the list. (In other words, the first row stays in place, regardless of its first letter).⏎When you

  4. #4 peak

    everything. Definition of terms that create your total calorie burn:⏎- RMR=resting metabolic rate-the amount of calories you need to exist without any activity-ie:bedrest⏎- TEA-thermic effect of activity-this, obviously,

  5. #5 peak

    you like. If you do find the new bulb to not give off enough light replace the new one with a higher “wattage” bulb. In other words instead of using a 820 lumen bulb to replace a 6

  6. #6 peak

    tool can prove to be very handy in the conversion of Arabic to Roman numerals. The tool basically works on a very simple concept of format conversion. In other words you just have to fill in the value of Arabic numeral and it will show 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: the book and the magazine "R. M. R. 3" and "R. M. R. 4" (and their subas) in the field of art and design. The design of these items is also a

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

    the piezoelectric transducers point to their applicability to wide range of dynamic pressure field measurements

  2. #2 peak

    5 section were instrumented with custom-made pressure transducers using polyvinylidene fluoride (PVDF) piezoelectric polymer film. These transducers provided in situ measurement of impulse pressures. These

  3. #3 peak

    Livermore National Laboratory researchers said their new tool enhances the way computer chips, LEDs and transistors are build.Commonly used piezo-electric speakers, such as those found in a cell phone, operate at low frequencies that

  4. #4 peak

    the Chilean salt flats, the Salar de Atacama — have been photographed from the International Space Station.⏎CALIPATRIA — There’s a lot to see at Simbol Materials’

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. L5:1847 suppresses
  2. L4:1559 suppresses
  3. L5:1424 excites
  4. L1:506 suppresses
  5. L5:1802 weakly suppresses
  6. L4:1498 weakly excites
  7. L4:1609 weakly excites
  8. L3:1847 weakly excites
  9. L3:2584 weakly suppresses
  10. L0:261 weakly excites
  11. L3:2565 weakly excites
  12. L4:2990 weakly suppresses
  13. L3:2135 weakly excites
  14. L1:2179 weakly suppresses
  15. L5:1870 weakly suppresses
  16. L4:301 weakly suppresses
  17. L0:2468 weakly suppresses
  18. L5:1161 weakly suppresses
  19. L1:379 weakly excites
  20. L2:2357 weakly excites
  21. L4:2018 weakly excites
  22. L2:2726 weakly excites 'by' after 'characterized' or 'marked'
  23. L2:2108 weakly suppresses
  24. L5:642 weakly suppresses

Feeds later layers

  1. L9:1921 excites
  2. L9:1624 weakly excites
  3. L13:2292 weakly excites opening parenthesis for a gloss or abbreviation
  4. L13:1216 weakly suppresses
  5. L16:2400 weakly suppresses
  6. L10:603 weakly suppresses
  7. L10:1142 weakly excites
  8. L12:2117 weakly suppresses
  9. L13:895 weakly excites
  10. L11:991 weakly suppresses
  11. L12:3032 weakly excites
  12. L8:2816 weakly excites
  13. L13:176 weakly suppresses
  14. L20:463 weakly suppresses
  15. L7:43 weakly excites
  16. L8:1707 weakly excites
  17. L20:2274 weakly excites
  18. L14:2610 weakly excites
  19. L12:1241 weakly excites
  20. L14:1527 weakly suppresses
  21. L11:2569 weakly excites
  22. L26:1388 weakly suppresses
  23. L26:697 weakly excites
  24. L10:2933 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.

3 values between -1.59 and -1.451 values between -1.31 and -1.175 values between -1.17 and -1.034 values between -1.03 and -0.893 values between -0.89 and -0.7614 values between -0.76 and -0.6215 values between -0.62 and -0.4850 values between -0.48 and -0.34112 values between -0.34 and -0.20589 values between -0.20 and -0.062001 values between -0.06 and 0.08848 values between 0.08 and 0.22200 values between 0.22 and 0.3641 values between 0.36 and 0.509 values between 0.50 and 0.642 values between 0.64 and 0.781 values between 0.78 and 0.921 values between 0.92 and 1.062 values between 1.20 and 1.341 values between 1.34 and 1.472 values between 1.47 and 1.611 values between 1.75 and 1.891 values between 1.89 and 2.031 values between 2.03 and 2.171 values between 3.84 and 3.98offnegativepositiveoften

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. _endian
  2. 网首页
  3. occer
  4. oned
  5. mandated
  6. anytime
  7. quam
  8. 前所
  9. alyzer
  10. oner

Suppresses

  1. _',
  2. *",
  3. ?',
  4. _",
  5. $",
  6. *',
  7. ',{
  8. ervative
  9. ",(
  10. ?",