Neuron Explorer

Qwen3-0.6B

'-based' after a place name

A neuron in layer 12 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 lights up on the word 'based' after a place, as in 'a London-based company' or 'a California-based researcher'. The few negative hits are on 'of' in phrases like 'some forms of'. Turning it up did not make the text talk about places or companies.

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

    from Israel.⏎Get exclusive in-depth reports from Israel⏎Connect with Israel, right from your home.⏎Lift up the voice of truth and hope⏎Support Jerusalem-based Zionist journalism.⏎Already a member? Login here.<|endoftext|>The fractured rock surfaces were

  2. #2 peak

    David Dubin⏎- LENS Videos⏎- Contact Us⏎Archive for author David Dubin⏎Los Angeles based LENS Neurofeedback provider and blogger.⏎More posts by David Dubin⏎Dr

  3. #3 peak

    the valves was unable to supply the new ones in the time required.⏎In response, Cristian Fracassi, CEO of Isinnova, a Brescia-based engineering firm, used 3D printing to meet the hospital’s demands and,

  4. #4 peak

    Gender Impacts of Mobile Money’ written by US-based economists Tavneet Suri and William Jack, and published in the prestigious

  5. #5 peak

    RiverWatch, an Austria-based NGO. “I sometimes think the western countries that are financially supporting this degradation process

  6. #6 peak

    since 1988, but it wasn’t confirmed until Charles J. Moore, a California-based sea researcher, stumbled upon the area in 1997 on his way

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 last week, we were in the state of New York, and I had just gone through the last week, and I had just gone through the state of New York, and I had just gone through the state of New York, and

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

    effect from some medication/antibiotics, so always read your pharmacy insert when getting prescriptions⏎- You have diabetes⏎- You are pregnant⏎- Some forms of contraception such as spermicides

  2. #2 peak

    is an inherited condition that implicates the specific enzyme responsible for the processing of bilirubin. Dubin-Johnson syndrome: This is an inherited form of chronic jaundice that prevents conjugated bilirubin from being secreted

  3. #3 peak

    or Center for Neuromodulation⏎What is Parkinson's disease?⏎Parkinson's disease (PD or, simply, Parkinson's) is the most common form of parkinsonism, a group of motor system disorders.⏎It is a slowly progressing

  4. #4 peak

    More than 1.4 billion adults worldwide are overweight and at increased risk for type 2 diabetes, heart disease, osteoarthritis, and certain forms of cancer.⏎âIf youâre overweight, losing just 5 to 1

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. L11:1174 suppresses
  2. L11:2497 suppresses
  3. L6:2326 suppresses
  4. L9:1909 excites
  5. L11:2207 suppresses "of" after a noun in trademark and title phrases
  6. L6:3070 excites
  7. L8:2755 suppresses
  8. L11:3015 suppresses
  9. L6:487 suppresses
  10. L8:1003 suppresses "the" or "of" before a noun phrase
  11. L11:2387 suppresses
  12. L6:1182 suppresses "of" after a noun phrase
  13. L6:1114 excites
  14. L6:1063 suppresses
  15. L5:363 excites
  16. L11:579 suppresses
  17. L1:1689 excites
  18. L6:1001 suppresses
  19. L9:2501 suppresses
  20. L6:1735 excites
  21. L9:2406 excites
  22. L5:262 weakly excites
  23. L7:714 weakly suppresses "of" after a noun introducing a category
  24. L10:486 weakly suppresses

Feeds later layers

  1. L13:1280 suppresses
  2. L16:2690 suppresses
  3. L13:602 suppresses
  4. L16:2196 excites
  5. L13:1515 excites "your" before doctor or health provider
  6. L17:1574 suppresses
  7. L14:2310 suppresses
  8. L13:486 excites
  9. L19:649 suppresses
  10. L14:1953 suppresses
  11. L19:1733 suppresses
  12. L17:275 suppresses
  13. L18:1099 excites
  14. L13:1906 excites
  15. L16:2612 excites
  16. L17:2733 excites
  17. L15:1507 suppresses
  18. L14:1550 suppresses
  19. L14:1192 suppresses
  20. L16:2530 excites
  21. L13:2135 suppresses
  22. L14:2044 suppresses
  23. L14:2107 excites
  24. L18:2090 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 -1.99 and -1.901 values between -1.90 and -1.804 values between -1.42 and -1.332 values between -1.33 and -1.231 values between -1.23 and -1.144 values between -1.14 and -1.044 values between -1.04 and -0.9510 values between -0.95 and -0.8511 values between -0.85 and -0.7615 values between -0.76 and -0.6624 values between -0.66 and -0.5732 values between -0.57 and -0.4750 values between -0.47 and -0.3898 values between -0.38 and -0.28150 values between -0.28 and -0.19380 values between -0.19 and -0.091400 values between -0.09 and 0.011143 values between 0.01 and 0.10325 values between 0.10 and 0.20120 values between 0.20 and 0.2960 values between 0.29 and 0.3916 values between 0.39 and 0.4815 values between 0.48 and 0.586 values between 0.58 and 0.676 values between 0.67 and 0.777 values between 0.77 and 0.862 values between 0.86 and 0.967 values between 0.96 and 1.051 values between 1.05 and 1.153 values between 1.15 and 1.243 values between 1.24 and 1.342 values between 1.34 and 1.432 values between 1.43 and 1.531 values between 1.53 and 1.621 values between 1.62 and 1.721 values between 1.72 and 1.81offnegativepositiveoften

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. quier
  2. ailer
  3. 来电
  4. 着他
  5. езн
  6. aviolet
  7. erable
  8. AppModule
  9. WISE
  10. odox

Suppresses

  1. ição
  2. --,
  3. iddle
  4. ights
  5. innen
  6. iores
  7. 年后
  8. iston
  9. DK
  10. _COLUMNS