Neuron Explorer

Qwen3-0.6B

opening bracket before an acronym

A neuron in layer 8 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

Fires on the opening bracket in writing like "Global Positioning System (GPS)", where a short form is about to be given for a longer name. Turning it up made the model drift into maths and generic text rather than acronyms.

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

    - Reduce cholesterol and triglyceride levels⏎- Lower blood pressure⏎- Improve memory and cognition⏎What about cinnamon and HIV⏎Acquired immunodeficiency syndrome (AIDS) is caused by HIV and HIV-1 protease plays an essential

  2. #2 peak

    ⏎- SOA services have self-describing interfaces in platform-independent XML documents. Web Services Description Language (WSDL) is used to describe the services.⏎- SOA services connects with

  3. #3 peak

    the Great Lakes. Research on potential tagging methods and movement of round gobies is scant. We explored the use of marking round gobies with passive integrated transponder (PIT) tags in order to determine the effects of tagging on growth and mortality

  4. #4 peak

    stop the gateway server from changing the messages or switching off the DKIM authentication.⏎What does SPF record mean?⏎It is highly recommended to have SPF (Sender Policy Framework) record for your domain, which is a type of DNS record that categorizes

  5. #5 peak

    another mother⏎The concept of digital twins, particularly when combined with MSPC, is relatively new, yet a well-constructed system based on Process Analytical Technology (PAT) will leverage both of these concepts, and as a result maximize your process

  6. #6 peak

    fecal coliform (FC) bacteria concentrations tend to vary with season because of seasonal variations in climatic factors. However, El Niño -Southern Oscillation (ENSO) events may affect the extent and patterns of seasonality in FC levels

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: 12, which is 10, and 11. I have used the Google and the Apple. If you have to use the Google. If you have to use the Google. If you have to use the Google.

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

    is a big part of getting your belly healthy again. As you can see in the image above, the TA is a thick layer of muscle. The transverse abdominal muscle wraps around the torso from front to back and the muscle fibers of the TA

  2. #2 peak

    bronze statue embodies the “development of Marxism” and shows respect to the revolutionary philosopher.⏎“Marx has significantly promoted the development of human society. The Communist Party of China has adopted Marx’s ideas, developing Marxism and using it in real life ...

  3. #3 peak

    Knowledge Base⏎NCSL Staff Contact⏎Revised June 2012⏎By Jacquelyn Pless⏎In recent years, technological advances in hydraulic fracturing and horizontal drilling have led to dramatic growth in natural gas development

  4. #4 peak

    , has focused efforts on ensuring that the practice is well regulated, transparent, and protects public health and the environment.⏎Although a number of federal regulations govern the hydraulic fracturing process, states have regulatory primacy on this issue. Knowledge of local ge

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. L7:810 suppresses
  2. L6:979 excites
  3. L0:10 weakly excites
  4. L4:556 weakly suppresses
  5. L4:2301 weakly excites
  6. L7:2415 weakly excites
  7. L3:856 weakly excites
  8. L5:1512 weakly excites
  9. L2:4 weakly suppresses
  10. L5:2398 weakly suppresses
  11. L3:2405 weakly excites
  12. L1:1253 weakly suppresses
  13. L1:892 weakly excites
  14. L1:72 weakly suppresses
  15. L1:2243 weakly excites
  16. L1:352 weakly suppresses
  17. L3:256 weakly excites
  18. L5:67 weakly excites
  19. L4:877 weakly excites
  20. L1:1150 weakly suppresses
  21. L1:2658 weakly suppresses
  22. L0:1914 weakly suppresses
  23. L1:2068 weakly excites
  24. L1:913 weakly suppresses

Feeds later layers

  1. L13:2722 suppresses
  2. L11:1471 suppresses
  3. L18:1308 suppresses opening parenthesis before an acronym
  4. L15:1146 suppresses
  5. L11:838 excites
  6. L18:2674 suppresses
  7. L12:570 excites
  8. L12:1303 weakly suppresses
  9. L16:2621 weakly suppresses
  10. L12:50 weakly suppresses
  11. L17:1710 weakly suppresses
  12. L15:762 weakly excites
  13. L13:298 weakly excites
  14. L16:1331 weakly excites
  15. L19:1860 weakly suppresses
  16. L16:2572 weakly suppresses
  17. L12:1902 weakly excites
  18. L16:1734 weakly suppresses
  19. L14:2023 weakly suppresses
  20. L17:2073 weakly excites
  21. L12:1783 weakly excites
  22. L14:3049 weakly excites
  23. L22:154 weakly excites
  24. L9:102 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 -2.97 and -2.841 values between -2.21 and -2.093 values between -1.96 and -1.831 values between -1.83 and -1.711 values between -1.71 and -1.582 values between -1.33 and -1.203 values between -1.20 and -1.089 values between -1.08 and -0.957 values between -0.95 and -0.8210 values between -0.82 and -0.7017 values between -0.70 and -0.5736 values between -0.57 and -0.4454 values between -0.44 and -0.32152 values between -0.32 and -0.19578 values between -0.19 and -0.061838 values between -0.06 and 0.06763 values between 0.06 and 0.19279 values between 0.19 and 0.32101 values between 0.32 and 0.4430 values between 0.44 and 0.577 values between 0.57 and 0.694 values between 0.69 and 0.822 values between 0.82 and 0.952 values between 0.95 and 1.071 values between 1.20 and 1.331 values between 1.33 and 1.452 values between 1.45 and 1.581 values between 1.83 and 1.962 values between 1.96 and 2.08offnegativepositiveoften

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. aran
  2. ATO
  3. ';";⏎
  4. чтобы
  5. colm
  6. 对公司
  7. zelf
  8. ====
  9. đốc
  10. sais

Suppresses

  1. elpers
  2. 好像
  3. .mix
  4. 似乎是
  5. wand
  6. estroy
  7. .EXIT
  8. olest
  9. NSE
  10. update