Neuron Explorer

Qwen3-0.6B

comma after an introductory phrase

A neuron in layer 13 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 at the point right after an opening phrase such as "Aside from the standard testing" or "While doctors treat illnesses", just before the main clause starts. The steering results did not show a 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

    for making special components.⏎Tips and Warnings⏎Don’t try using a lathe unless you know exactly what you’re doing. Aside from the precautions mentioned earlier you should always roll the sleeves above the elbows.⏎Stop the lathe before adjusting any

  2. #2 peak

    iromancy, however, is kind of vague, with each ancient culture offering different variations in dream interpretation. And aside from the dream meanings advocated by our ancestors, some great psychoanalysts have also offered

  3. #3 peak

    ological properties, differences in the relative mechanical properties can be determined during the acquisition of an image and subsequently used as an indirect chemical probe. Beyond analysis under ambient conditions, atomic force microscopes can easily be outfitted for detailed relative humidity and reactive gas

  4. #4 peak

    and other toxic substances build up in the blood, causing life-threatening complications shortly after birth.⏎While infants with citrullinemia type I appear normal at birth, within the first week of life, most will become lethargic and display poor

  5. #5 peak

    public health is the impact it can make. Public health is community health. Medicine is to the individual as public health is to the population. While doctors treat illnesses, public health professionals implement health promotion and health education.⏎Unlike medicine, there is no

  6. #6 peak

    saying.⏎Interestingly, I’ve been able to hold space more on the professional level while practicing optometry for 20 years. Aside from the standard clinical testing, I would sit and give my patients my full attention. I was always struck

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: When you have a problem to be solved, it is not only as the problem can be proved to be false or true or to be (as well as the rest of this) and then it is not a problem to be proved to

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

    > <H1>This is the Page Heading</H1> This is the first paragraph. <P> This is another paragraph, with

  2. #2 peak

    and captured by the U.S. Army 2nd Division in Korea, during the Korean War. A convoy of jeeps and trucks on the road. The trucks are loaded with supplies and soldiers. Soldiers on trucks unload rations,

  3. #3 peak

    one put a price on historic culture?”⏎While the answer seems obvious, the reality from a fiscal viewpoint is that this rare community treasure does not sustain itself financially.⏎From The Start⏎The original Rancho site was 27,

  4. #4 peak

    designed to provide cosmetic chemists and formulators with ideas and advice to solve formulating challenges. To submit a question, e-mail FormulateThis@allured.com.⏎Question: I’m trying to understand cosmetic emulsion inversion. What actually happens

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. L12:71 excites
  2. L12:51 excites
  3. L11:2214 excites
  4. L11:567 excites
  5. L8:131 suppresses
  6. L12:2548 excites
  7. L6:1084 suppresses
  8. L12:959 weakly excites
  9. L12:1443 weakly suppresses opening parenthesis after a name or term
  10. L9:1458 weakly excites unclear: word completing a verb phrase
  11. L9:2181 weakly excites
  12. L8:70 weakly excites 'and' between stacked superlatives
  13. L12:2208 weakly excites
  14. L9:1374 weakly excites
  15. L10:1654 weakly suppresses opening parenthesis before an acronym
  16. L12:1254 weakly excites
  17. L9:283 weakly suppresses
  18. L11:1205 weakly suppresses
  19. L10:144 weakly suppresses
  20. L7:398 weakly suppresses
  21. L6:982 weakly excites
  22. L11:1116 weakly excites
  23. L8:1477 weakly excites
  24. L6:1652 weakly excites

Feeds later layers

  1. L15:1664 excites
  2. L15:252 weakly suppresses
  3. L14:2204 weakly suppresses
  4. L17:2269 weakly suppresses
  5. L17:2864 weakly suppresses
  6. L15:2114 weakly excites
  7. L15:220 weakly suppresses
  8. L19:1689 weakly excites
  9. L16:2665 weakly excites
  10. L16:520 weakly suppresses
  11. L20:2854 weakly excites 'official' and government institution words
  12. L15:36 weakly suppresses
  13. L16:1121 weakly suppresses
  14. L24:2998 weakly excites
  15. L20:1727 weakly excites
  16. L24:1361 weakly excites
  17. L18:1169 weakly suppresses
  18. L14:2616 weakly excites
  19. L14:578 weakly excites
  20. L27:38 weakly excites
  21. L16:2865 weakly suppresses
  22. L14:53 weakly excites
  23. L15:2628 weakly suppresses
  24. L23:748 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 -1.38 and -1.251 values between -1.25 and -1.122 values between -1.12 and -1.002 values between -1.00 and -0.877 values between -0.87 and -0.7415 values between -0.74 and -0.6125 values between -0.61 and -0.4864 values between -0.48 and -0.36130 values between -0.36 and -0.23456 values between -0.23 and -0.101946 values between -0.10 and 0.03867 values between 0.03 and 0.16223 values between 0.16 and 0.2957 values between 0.29 and 0.4127 values between 0.41 and 0.5422 values between 0.54 and 0.6724 values between 0.67 and 0.809 values between 0.80 and 0.933 values between 0.93 and 1.066 values between 1.06 and 1.187 values between 1.18 and 1.312 values between 1.31 and 1.441 values between 1.44 and 1.573 values between 1.57 and 1.701 values between 1.70 and 1.831 values between 1.83 and 1.951 values between 1.95 and 2.081 values between 2.08 and 2.211 values between 2.21 and 2.342 values between 2.98 and 3.111 values between 3.62 and 3.75offnegativepositiveoften

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. star
  2. combat
  3. lesi
  4. oe
  5. Stars
  6. lt
  7. conds
  8. oi
  9. stars
  10. gear

Suppresses

  1. paring
  2. ışı
  3. -floating
  4. IFA
  5. ||(
  6. retrieval
  7. 这一点
  8. Went
  9. .IsDBNull
  10. ritten