Neuron Explorer

Qwen3-0.6B

'on' or 'after' before a measured topic

A neuron in layer 6 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 small connecting words like 'on', 'after' or 'in' that come right before the subject being studied or assessed, such as 'assessed based on' or 'scouts their fields after'. Turning it up did not give a clear story in the output.

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

    43 ELSA-Brazil participants with an average age of 51.9 years who had not been diagnosed with heart disease. Subjects were assessed based on body mass index, hypertension, diabetes, cholesterol levels, smoking, physical activity,

  2. #2 peak

    evidence substantiation.⏎You will need to understand the sentence’s intention and keep an eye out for brevity and succinctness.⏎You will be quizzed on all the usual suspects: commas, semicolons and colons; subject-

  3. #3 peak

    Dimensional Analysis content provides you with well-rounded coverage.⏎- Pretest and post-test help identify strengths and weaknesses in competency of basic math before and assess your comprehension after Unit One:

  4. #4 peak

    try and cd⏎into a directory, it will only try and change to the path specified from the current directory. If you spend a⏎lot of time working in specific folders, it might be useful to assume that your search path for⏎cd

  5. #5 peak

    using color as an indicator is an unsafe practice, as one part of the chicken may be done, but other pieces might not. When checking doneness of poultry, make sure to put the

  6. #6 peak

    cooler soil temperatures, which delays soil warm up, planting and emergence. These conditions make plants more susceptible to insects."⏎Davis suggests that growers need to scout their fields after planting to see how well their seed treatments are working. She advises growers to watch

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: this is a list of possible solutions. if the answer is not found, then I will have to make sure that the answer is correct. The answer should be in the form of a list of possible solutions. **Possible solutions:** **

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

    in an effort to understand the intergeneration effects of trauma specifically the relationship between a parent and child's ACE scores. Thousands of families across the country will be screened to examine ACE screening results, child social and emotional development and parent functioning; as well

  2. #2 peak

    of our everyday conversation. Now scientists can explain why: Doing so stimulates the brain the same way sex, food, and cash do. Researchers scanned subjects' brains and found that parts linked to those pleasures lit up when subjects chatted about themselves.⏎

  3. #3 peak

    the transport something strange started to happen, the juvenile males were becoming aggressive and violently killing other animals in the new park. It was chaos.⏎After much research, it was discovered that the absents of the mature bull elephants was the culprit.

  4. #4 peak

    and mental health from a random adult sample that’s representative of the U.S. civilian population. After sifting through 135,095 respondents, they found that almost 14% reported use of one of the classical psyched

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. L2:2281 weakly excites
  2. L3:278 weakly excites
  3. L5:911 weakly excites
  4. L1:354 weakly suppresses
  5. L1:2821 weakly suppresses
  6. L2:319 weakly suppresses 'of' after 'pattern' or 'patterns'
  7. L5:1213 weakly suppresses
  8. L4:2440 weakly excites
  9. L5:102 weakly excites
  10. L2:2293 weakly suppresses
  11. L3:919 weakly excites
  12. L4:1410 weakly excites
  13. L0:1296 weakly excites
  14. L5:91 weakly excites
  15. L5:455 weakly suppresses
  16. L1:1158 weakly suppresses
  17. L5:2854 weakly suppresses
  18. L5:723 weakly suppresses
  19. L4:829 weakly excites
  20. L4:2645 weakly excites
  21. L4:316 weakly suppresses
  22. L2:1287 weakly suppresses
  23. L5:1639 weakly suppresses
  24. L3:103 weakly excites

Feeds later layers

  1. L7:2082 suppresses
  2. L10:2186 suppresses
  3. L12:2761 excites
  4. L17:2995 suppresses
  5. L13:1366 suppresses
  6. L7:2179 suppresses
  7. L15:2512 weakly suppresses
  8. L11:1660 weakly suppresses
  9. L16:1672 weakly suppresses
  10. L8:3005 weakly suppresses
  11. L7:2955 weakly excites
  12. L8:1454 weakly suppresses
  13. L24:1941 weakly suppresses
  14. L11:1235 weakly excites
  15. L9:1044 weakly suppresses
  16. L20:2396 weakly suppresses
  17. L10:1593 weakly suppresses
  18. L14:1467 weakly suppresses
  19. L7:893 weakly excites
  20. L10:877 weakly suppresses
  21. L17:2885 weakly suppresses
  22. L20:2224 weakly suppresses
  23. L16:2732 weakly suppresses
  24. L12:1812 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.

1 values between -1.88 and -1.791 values between -1.79 and -1.691 values between -1.50 and -1.412 values between -0.94 and -0.854 values between -0.85 and -0.756 values between -0.75 and -0.668 values between -0.66 and -0.569 values between -0.56 and -0.4717 values between -0.47 and -0.3757 values between -0.37 and -0.28163 values between -0.28 and -0.19460 values between -0.19 and -0.091311 values between -0.09 and 0.001081 values between 0.00 and 0.10469 values between 0.10 and 0.19192 values between 0.19 and 0.2971 values between 0.29 and 0.3824 values between 0.38 and 0.477 values between 0.47 and 0.575 values between 0.57 and 0.669 values between 0.66 and 0.764 values between 0.76 and 0.852 values between 0.85 and 0.941 values between 0.94 and 1.041 values between 1.23 and 1.321 values between 1.51 and 1.601 values between 1.79 and 1.89offnegativepositiveoften

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. 轴
  2. owment
  3. around
  4. 动态
  5. histó
  6. roleum
  7. 角
  8. rotary
  9. ido
  10. PIO

Suppresses

  1. conclusion
  2. RESULTS
  3. onica
  4. conclusions
  5. adders
  6. 判定
  7. conclude
  8. findFirst
  9. _find
  10. ฬ