Neuron Explorer

Qwen3-0.6B

capital F starting a word mid-sentence

A neuron in layer 25 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 a capital F that begins a name or acronym, such as FPGAs, FANUC, FDM or Fabrik, usually right after another word. Steering did not produce any clear theme.

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

    significantly in their architecture, programming, and functionality. In this article, we will delve into the main differences between microcontrollers and FPGAs, as well as their advantages and disadvantages.⏎What is a Microcontroller

  2. #2 peak

    ? That’s exactly what the duo of Maxim Tsoy and Wilbert Swinkels has developed. Inspired by other DIY cube solvers comprised of LEGO and Fischertechnik, the Makers brought their creation

  3. #3 peak

    FPGA’s hardware components to perform specific functions.⏎Microcontrollers are designed to handle simple tasks, making them a popular choice in low-cost, low-power embedded systems. FPGAs, on the other hand, are designed to handle complex logic functions that

  4. #4 peak

    nece between the servo spindle and variable frequency spindle?⏎At present, the primary control of the FANUC system’s spindle motor has two interfaces: analog (DC0 ~

  5. #5 peak

    in a repeating event, so the period is the reciprocal of the frequency.⏎Deutsch: Fabrik / Español: Fábrica / Português: Fábrica / Français: Usine / Italiano: Fabbrica⏎

  6. #6 peak

    CNC machines, we had to keep the geometry simple because complex shapes would greatly increase manufacturing costs," says WHOI Supervising Engineer Bob McCabe." Then we heard about FDM and its ability to produce polycarbonate or ABS parts with the strength and

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 open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open open

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

    requires collection and integration of complex multi-disciplinary datasets. The first presentation will describe the data management system being developed by PROTECT to integrate, manage, analyze, and relate environmental, demographic, exposure biomarkers, and birth outcome data. The discussion

  2. #2 peak

    7 release of the JavaServer Pages 1.2 specification (⏎- The new specification includes better mechanisms for accessing and describing XML data than earlier versions.⏎- Standard JSP tag libraries support XML based data more efficiently and effectively.⏎- Lots of third

  3. #3 peak

    , results, hypotheses and opinions, lie at the very heart of scientific process. In this context, Open Science is the growing global movement to make scientific research and data accessible to all.⏎Open Science has the potential to significantly increase scientific collaboration and discovery

  4. #4 peak

    (GIS) is a framework for gathering, managing, and analyzing data. Incorporating GIS into the disaster response framework has improved planning, response

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. L4:1165 suppresses
  2. L7:2930 weakly suppresses
  3. L19:1313 weakly excites
  4. L22:500 weakly excites
  5. L3:857 weakly suppresses
  6. L15:1767 weakly excites
  7. L20:461 weakly suppresses
  8. L4:2988 weakly suppresses
  9. L11:2097 weakly suppresses
  10. L20:1255 weakly excites
  11. L21:832 weakly excites
  12. L3:688 weakly suppresses
  13. L24:1714 weakly excites
  14. L15:847 weakly suppresses
  15. L19:1231 weakly excites
  16. L20:1359 weakly suppresses
  17. L7:2721 weakly excites
  18. L4:1853 weakly excites
  19. L20:45 weakly excites
  20. L14:909 weakly suppresses
  21. L13:1509 weakly suppresses
  22. L8:91 weakly excites
  23. L22:1510 weakly excites
  24. L13:1103 weakly excites

Feeds later layers

  1. L26:1692 weakly suppresses
  2. L26:2740 weakly excites
  3. L27:788 weakly excites
  4. L27:2880 weakly excites
  5. L26:1823 weakly excites
  6. L26:1709 weakly excites
  7. L27:2473 weakly excites
  8. L27:3005 weakly excites
  9. L27:2952 weakly excites
  10. L27:803 weakly excites
  11. L27:2664 weakly excites
  12. L27:1138 weakly excites
  13. L27:770 weakly excites
  14. L26:1999 weakly suppresses
  15. L27:2738 weakly excites unclear: fragments of split words
  16. L27:2022 weakly excites
  17. L27:1003 weakly suppresses
  18. L26:1275 weakly suppresses
  19. L26:1186 weakly excites
  20. L26:1543 weakly excites
  21. L26:1346 weakly excites
  22. L27:1899 weakly suppresses
  23. L27:757 weakly excites
  24. L26:3066 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 -33.16 and -31.891 values between -25.55 and -24.281 values between -24.28 and -23.012 values between -23.01 and -21.742 values between -21.74 and -20.471 values between -20.47 and -19.201 values between -19.20 and -17.943 values between -17.94 and -16.674 values between -16.67 and -15.402 values between -15.40 and -14.136 values between -14.13 and -12.866 values between -12.86 and -11.597 values between -11.59 and -10.3325 values between -10.33 and -9.0643 values between -9.06 and -7.7957 values between -7.79 and -6.5278 values between -6.52 and -5.2593 values between -5.25 and -3.98155 values between -3.98 and -2.72279 values between -2.72 and -1.45903 values between -1.45 and -0.182067 values between -0.18 and 1.09112 values between 1.09 and 2.3628 values between 2.36 and 3.6315 values between 3.63 and 4.906 values between 4.90 and 6.164 values between 6.16 and 7.431 values between 7.43 and 8.703 values between 8.70 and 9.971 values between 9.97 and 11.241 values between 16.31 and 17.58offnegativepositiveoften

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. 损
  2. lies
  3. .loss
  4. 损失
  5. 喘
  6. ache
  7. /root
  8. 懒
  9. losses
  10. 丧失

Suppresses

  1. казать
  2. ector
  3. *[
  4. otion
  5. ступил
  6. cí
  7. {[
  8. cé
  9. oscope
  10. tero