Neuron Explorer

Qwen3-0.6B

'for' after preparation or purpose nouns

A neuron in layer 1 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 the word "for" when it follows words like "preparation", "preparing", or "earmarked", where the next thing is a purpose or target. Its negative side is weak and shows up on "for" in loop or quantifier uses. Turning it up gave text that changed topic, so the steering doesn't clearly show its role.

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

    cut a large tree shape out of craft paper and affix it to the dining room wall in preparation for thanksgiving. Then, during the meal, each family member and friend turn over

  2. #2 peak

    age of eight weeks, dogs scheduled for surgery, and people with pre-existing anemia mustn't be given garlic.⏎

  3. #3 peak

    in March 2023, which brought together<|endoftext|>Before I go further, I want to emphasize that I believe that school is not just about preparing students for the future. It's about helping them manage their now. I think we often

  4. #4 peak

    15 inches of rain over a 72-hour period.⏎Readiness, Response and Recovery⏎With a forecast for heavy rainfall, the District began preparations for Tropical Storm Isaac in the days before it impacted South Florida. Water levels in regional

  5. #5 peak

    Curtiss and purchase a plane for $7,000 ($157,000 in today

  6. #6 peak

    breast—near the turkey's crop—is known as breast sponge. Gobblers develop the fatty layer of breast sponge in late winter as they feed heavily in preparation for breeding season. The fat deposit enables gobblers to have energy to spend little time

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 have been thinking about lately, and I have been thinking about lately. So I have been thinking about lately. So I have been thinking about lately. So I have been thinking about lately. So I have been thinking about lately. So

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

    if n> 10⏎> 3 print "This might take a while..."⏎> 4 for i = 1 to n⏎> 5 for j = 1 to i⏎> 6 print

  2. #2 peak

    for all types of wrens. These birds can have substantial broods, often from

  3. #3 peak

    for all the battles that wound up half-won, it was a romantic time in our

  4. #4 peak

    langue was responsible⏎for the financing of the building, and it was supposed to build the auberge

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. L0:1301 suppresses
  2. L0:178 suppresses
  3. L0:1224 suppresses
  4. L0:1 weakly suppresses
  5. L0:9 weakly suppresses
  6. L0:1960 weakly excites unclear: mixed medical and harsh words
  7. L0:3031 weakly excites
  8. L0:33 weakly excites
  9. L0:1824 weakly excites
  10. L0:1148 weakly suppresses
  11. L0:2031 weakly suppresses
  12. L0:1512 weakly excites
  13. L0:492 weakly excites
  14. L0:154 weakly suppresses
  15. L0:2538 weakly suppresses
  16. L0:81 weakly suppresses
  17. L0:97 weakly excites
  18. L0:1796 weakly excites
  19. L0:2443 weakly suppresses
  20. L0:1378 weakly excites
  21. L0:3058 weakly excites
  22. L0:908 weakly excites
  23. L0:2072 weakly excites
  24. L0:3 weakly suppresses

Feeds later layers

  1. L10:1648 weakly excites
  2. L14:2150 weakly suppresses
  3. L24:2437 weakly excites
  4. L11:2269 weakly suppresses
  5. L14:3052 weakly suppresses
  6. L3:2469 weakly excites
  7. L7:2815 weakly excites
  8. L15:692 weakly suppresses
  9. L3:2710 weakly excites
  10. L10:800 weakly excites
  11. L7:1855 weakly excites
  12. L7:2427 weakly suppresses word before 'type 2 diabetes' or a condition
  13. L26:122 weakly excites
  14. L11:374 weakly excites
  15. L11:2280 weakly suppresses
  16. L6:1045 weakly excites
  17. L7:1460 weakly suppresses
  18. L18:967 weakly excites
  19. L6:1660 weakly suppresses
  20. L24:289 weakly excites
  21. L13:822 weakly suppresses
  22. L8:2918 weakly suppresses
  23. L18:552 weakly excites
  24. L12:1474 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 -0.44 and -0.362 values between -0.36 and -0.2930 values between -0.29 and -0.21125 values between -0.21 and -0.14494 values between -0.14 and -0.061973 values between -0.06 and 0.011004 values between 0.01 and 0.09186 values between 0.09 and 0.1657 values between 0.16 and 0.2310 values between 0.23 and 0.316 values between 0.31 and 0.383 values between 0.38 and 0.461 values between 0.46 and 0.535 values between 0.53 and 0.612 values between 0.61 and 0.681 values between 0.83 and 0.901 values between 0.90 and 0.981 values between 0.98 and 1.051 values between 1.13 and 1.201 values between 1.27 and 1.351 values between 1.50 and 1.571 values between 1.87 and 1.941 values between 2.24 and 2.311 values between 2.46 and 2.54offnegativepositiveoften

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. 什么样的
  2. -side
  3. etched
  4. -sided
  5. médico
  6. ей
  7. ós
  8. :\/\/
  9. pared
  10. (),"

Suppresses

  1. commod
  2. Ups
  3. Institutes
  4. ButtonItem
  5. я
  6. ji
  7. Channels
  8. issa
  9. CHANNEL
  10. nty