Neuron Explorer

Qwen3-0.6B

inside organisation and product names

A neuron in layer 17 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 a word in the middle of a proper name, such as the first part of an institution, product or event name, just before the name finishes. Turning it up did not produce a clear theme; the outputs became odd and unrelated.

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

    Wisconsin-Madison are doing something like that (although, of course, the weights don't weigh anything) by creating a virtual reality simulator called the Cave Automatic Virtual Environment, or just CAVE, that works by tracking your muscle movements. Users

  2. #2 peak

    , let your child paint and decorate their egg carton. Your child will love this!⏎I stumbled upon this Montessori-inspired sound matching activity over on Living Montessori

  3. #3 peak

    some people out there who come up with creative ideas to close the gap.⏎To celebrate the 40 year aniversary of the first moon landing Google released Google Moon. Now, a

  4. #4 peak

    of<|endoftext|>How do you bring an endangered plant species back from the brink of extinction? The answer might be found in zoo animals.⏎That’s the inspiration for Chicago Botanic Garden scientist Jeremie Fant’s latest research. Fant, a molecular ecologist

  5. #5 peak

    therefore rapidly position Canadian epigenetics research at the forefront of the International Human Epigenome Consortium (IHEC). IHEC seeks to coordinate human ep

  6. #6 peak

    021⏎Interpretation of change:⏎A high or increasing life expectancy is desirable. An increase from the baseline year is an improvement.⏎Name: ABS Life Tables for Aboriginal and Torres Strait Islander Australians⏎Provider name: Australian Bureau of

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 more you have to do with the things that we can do without, we can start to take a more than just 'something' - as in, we can have more things, by having more things that we can take things away from

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

    companies, and other institutions, and they pass on those costs to customers.⏎A foundation-supported initiative allows Rwandan farmers to access markets<|endoftext|>The small rufous-sided pygmy tyrant is a member of the family Tyrannidae,

  2. #2 peak

    . The change was reflected in the second formant values of /ʉ/, which lowered significantly after three trainings. The results show that 9–11-year-old children benefit from passive auditory training in second language production learning regardless of whether or

  3. #3 peak

    stress and to resist disease.”⏎The most popular IF methods include:⏎- Alternate day fasting⏎- 5/2 fasting, which involves eating regularly for 5

  4. #4 peak

    -Grant; Pulmonary-system-disorders; Dust-control; Air-quality-control; Measurement-instrumentation; Measurement

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. L16:847 strongly excites
  2. L14:817 strongly excites
  3. L16:1064 strongly excites
  4. L14:732 strongly excites
  5. L15:790 strongly excites
  6. L14:459 excites
  7. L15:1057 excites
  8. L15:1460 excites
  9. L15:1701 excites
  10. L10:856 suppresses
  11. L15:629 excites
  12. L15:1022 excites
  13. L15:658 excites
  14. L6:189 suppresses
  15. L16:627 excites
  16. L15:1395 suppresses
  17. L15:661 suppresses
  18. L16:495 suppresses
  19. L16:1206 suppresses
  20. L16:337 suppresses
  21. L10:383 excites
  22. L15:763 excites
  23. L14:402 excites
  24. L10:1274 excites

Feeds later layers

  1. L18:595 strongly excites
  2. L19:1860 strongly excites
  3. L18:543 strongly excites
  4. L20:494 excites
  5. L18:1081 excites
  6. L20:2882 excites
  7. L18:318 suppresses
  8. L22:315 excites
  9. L18:167 excites
  10. L18:2835 excites first piece of a split name or term
  11. L18:276 excites
  12. L18:660 excites
  13. L18:279 excites
  14. L19:430 excites
  15. L19:2054 excites
  16. L18:970 suppresses
  17. L21:36 suppresses
  18. L21:31 suppresses
  19. L18:2217 excites
  20. L20:463 excites
  21. L19:224 weakly excites
  22. L18:518 weakly excites
  23. L21:39 weakly suppresses
  24. L18:121 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 -5.72 and -5.322 values between -4.12 and -3.725 values between -3.72 and -3.324 values between -3.32 and -2.922 values between -2.92 and -2.524 values between -2.52 and -2.126 values between -2.12 and -1.7213 values between -1.72 and -1.3347 values between -1.33 and -0.93430 values between -0.93 and -0.531202 values between -0.53 and -0.131316 values between -0.13 and 0.27264 values between 0.27 and 0.67144 values between 0.67 and 1.07106 values between 1.07 and 1.4766 values between 1.47 and 1.8766 values between 1.87 and 2.2750 values between 2.27 and 2.6748 values between 2.67 and 3.0727 values between 3.07 and 3.4716 values between 3.47 and 3.8628 values between 3.86 and 4.2615 values between 4.26 and 4.6610 values between 4.66 and 5.067 values between 5.06 and 5.465 values between 5.46 and 5.862 values between 5.86 and 6.264 values between 6.26 and 6.664 values between 6.66 and 7.066 values between 7.06 and 7.461 values between 7.86 and 8.252 values between 8.25 and 8.652 values between 9.45 and 9.853 values between 9.85 and 10.25offnegativepositiveoften

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. olio
  2. syndrome
  3. imonial
  4. ach
  5. heritance
  6. urses
  7. Syndrome
  8. tm
  9. ociety
  10. riors

Suppresses

  1. تا
  2. iese
  3. ppy
  4. athi
  5. ȩ
  6. _iso
  7. arkan
  8. rette
  9. 埃
  10. .Resume