Neuron Explorer

Qwen3-0.6B

unclear: mid-word splits in foreign or code text

A neuron in layer 8 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 odd word pieces, like the middle of a split word or a fragment inside a web address or a foreign word. When it was turned up, the model's writing drifted toward topics like school and technical help, and it got less coherent.

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

    years later<|endoftext|>Hand-coloured lithograph entitled; 'Dampfmaschinen Pferd worauf man in einer stunde von Paris nach Petersburg reiten kann' ('Steam engine horse on which one can

  2. #2 peak

    ulunda tamamlamıştır.⏎Yüksek Lisans Öğrenimine MSÜ Barbaros DEBİM Enstitüsünde devam etmekte aynı zamanda enerji sistem ve modelleme üzerine araştırm

  3. #3 peak

    those involved in educational leadership and policy development. This work is also valuable for those interested in the locally organized and interactionally achieved context of institutional work' - Discourse:

  4. #4 peak

    7075. . . http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0057075.<|endoftext|>Racism

  5. #5 peak

    Sustainability 2013, 5(7), 3006-3008; doi:10.3390/su5073006⏎I am honored to introduce this

  6. #6 peak

    yellow. The french blog Volcanic News Erutpion states it as being restless. http://volcansvanuatueruptionsgb.blogspot.fr/2014/03/08032014

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: I need to make a 200/300/300/300 configuration for a home-made 200/300/300 configuration. The first part is 200

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

    .⏎For intramuscular use only.⏎Recommended Dose And Schedule⏎The dose and schedule for Fluvirin is presented in Table 1.⏎TABLE 1 : Fluvirin Dose

  2. #2 peak

    enclosure is designed to be corrosion and UV resistant and strong enough to withstand falling trees. The electronics within the enclosure are shown in Figure 6.⏎Figure 5. Typical element enclosure.⏎Wind speed, wind direction, and temperature are also recorded at

  3. #3 peak

    The severity of fractures with large pieces of bone dislodged indicate that these were formed from violent blows in a standardised manner (Figure 1).⏎Figure 1: Top row, Cranium 20

  4. #4 peak

    , but every year the Internet packet transmission market could grow as shown in Figure 4-1.⏎Figure 4-1. The unit cost of Internet bandwidth drops, but the volume increases.⏎The result

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. L6:1601 excites
  2. L7:2701 weakly suppresses
  3. L2:67 weakly excites
  4. L3:23 weakly suppresses
  5. L5:1938 weakly suppresses
  6. L5:1319 weakly suppresses
  7. L2:2771 weakly excites
  8. L7:1972 weakly excites
  9. L1:168 weakly suppresses
  10. L2:2202 weakly excites
  11. L6:1102 weakly excites
  12. L2:1465 weakly excites
  13. L5:919 weakly suppresses
  14. L0:2304 weakly excites
  15. L4:1963 weakly excites
  16. L0:2769 weakly suppresses
  17. L7:3029 weakly suppresses
  18. L6:2364 weakly excites
  19. L4:245 weakly excites
  20. L6:2479 weakly excites
  21. L7:1970 weakly suppresses
  22. L3:2884 weakly excites
  23. L1:1272 weakly suppresses
  24. L2:2956 weakly suppresses

Feeds later layers

  1. L10:1874 suppresses
  2. L10:586 excites 'important' and degree words after 'how'
  3. L9:799 suppresses
  4. L19:1361 weakly excites
  5. L14:1498 weakly excites
  6. L20:810 weakly suppresses
  7. L9:288 weakly excites
  8. L13:1513 weakly excites
  9. L16:1173 weakly excites
  10. L25:1693 weakly suppresses
  11. L15:1955 weakly suppresses
  12. L16:1036 weakly excites
  13. L19:2756 weakly suppresses
  14. L23:1186 weakly excites
  15. L19:2575 weakly suppresses
  16. L14:827 weakly suppresses open bracket after a value or citation
  17. L18:104 weakly excites
  18. L21:1035 weakly suppresses
  19. L23:2273 weakly suppresses
  20. L19:45 weakly excites
  21. L13:1001 weakly suppresses
  22. L27:1244 weakly suppresses
  23. L11:687 weakly excites
  24. L20:2816 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 -1.13 and -1.091 values between -1.04 and -1.001 values between -0.86 and -0.822 values between -0.59 and -0.543 values between -0.54 and -0.504 values between -0.50 and -0.457 values between -0.45 and -0.417 values between -0.41 and -0.3629 values between -0.36 and -0.3239 values between -0.32 and -0.2784 values between -0.27 and -0.23118 values between -0.23 and -0.18223 values between -0.18 and -0.14312 values between -0.14 and -0.09405 values between -0.09 and -0.05565 values between -0.05 and -0.00605 values between -0.00 and 0.04520 values between 0.04 and 0.09372 values between 0.09 and 0.13243 values between 0.13 and 0.18137 values between 0.18 and 0.2391 values between 0.23 and 0.2749 values between 0.27 and 0.3238 values between 0.32 and 0.3620 values between 0.36 and 0.4118 values between 0.41 and 0.457 values between 0.45 and 0.503 values between 0.50 and 0.542 values between 0.54 and 0.591 values between 0.59 and 0.631 values between 0.63 and 0.68offnegativepositiveoften

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. anes
  2. 萦
  3. uada
  4. apel
  5. เอง
  6. 之势
  7. 羌
  8. ושר
  9. 都要
  10. ower

Suppresses

  1. :///
  2. "<?
  3. 读后
  4. reads
  5. boa
  6. ented
  7. ento
  8. ello
  9. ęp
  10. adays