Neuron Explorer

Qwen3-0.6B

comma after a conditional lead-in clause

A neuron in layer 14 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 on the comma that ends an "if" or "when" style opening phrase, such as "given enough time and sufficient access," or "once in place and properly utilized," just before the main point. Its negative side is the opposite: a comma after a "After treatment," or "After demolding," phrase. Turning it up did not produce clearly related writing.

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

    . This type of writing essays might not be popular with the crowd, as it is thought to be difficult to write persuasive essays, but if you are well prepared it could be one of the most enjoyable styles of writing for college.⏎You should write

  2. #2 peak

    the team will spend about 45 days in the Southern Ocean, most of which will be spent studying the ice. But, given enough time and sufficient access, they hope to catch a glimpse of the famous ship too. Only time will tell

  3. #3 peak

    ground with the clinker during manufacture. These cements produce concrete with improved resistance to freezing and thawing. When mixed with proper intensity and for an appropriate duration, such<|endoftext|>In this section⏎

  4. #4 peak

    for breathable oxygen. The mind will become burning off its capability to regulate breathing in, therefore. If heVersusshe has the ailment and presented suitable treatment, with the assistance of a health care provider, an be clinically determined. Treatment method

  5. #5 peak

    range in the late 1800s because of unregulated trapping. With a decline in the demand for beaver pelts, and with proper management, they became reestablished in much of their former range and are now common in

  6. #6 peak

    ; I'd sell a million books -- plus, the headlines would read "Philplait to Destroy the Earth!" which is awesome). If it could hit us, it would have an impact yield of at least 35,000

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: I still want to make sure I can get the system and the team still do something. I don't want to know if they are still there, just that I can get them. I don't want to do anything. I don't

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 too little bonding performance of the slurry, and the slurry will not be able to adhere tightly to the foamed plastic net. After demolding, it will have no strength and will easily disassemble. Hang up even if it

  2. #2 peak

    oprazole or omeprazole are prescribed in various combinations.⏎Finishing all medicines prescribed by your doctor is very important for successfully eliminating infection.⏎After treatment, children may sometimes complain of discomfort in their stomach and this can take several months to

  3. #3 peak

    was at the DDD exchange recently where we had the likes of Udi Dahan, Eric Evans and Scott Wlaschin on the panel. In a post event Q&A session I asked the panel - “Is microservices SOA renamed?

  4. #4 peak

    , and has the default set of MySQL accounts and no user data.⏎On Unix, initialize the data directory. See Section 2.10, “Postinstallation Setup and Testing”.⏎On Windows

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. L11:1569 suppresses
  2. L13:1569 weakly suppresses
  3. L4:2769 weakly suppresses
  4. L7:307 weakly suppresses
  5. L13:1173 weakly suppresses
  6. L11:1399 weakly suppresses
  7. L11:1225 weakly suppresses
  8. L7:747 weakly suppresses
  9. L3:2431 weakly excites
  10. L9:1192 weakly suppresses
  11. L12:2718 weakly excites
  12. L7:1964 weakly suppresses
  13. L1:3026 weakly suppresses
  14. L12:1872 weakly suppresses
  15. L12:238 weakly excites
  16. L12:1099 weakly suppresses
  17. L12:3000 weakly suppresses
  18. L13:1874 weakly excites
  19. L11:2112 weakly suppresses
  20. L12:534 weakly suppresses
  21. L12:2185 weakly suppresses
  22. L3:1167 weakly excites
  23. L12:179 weakly suppresses
  24. L2:2863 weakly suppresses

Feeds later layers

  1. L16:1470 suppresses
  2. L19:1863 excites
  3. L18:2592 weakly suppresses comma after an opening clause
  4. L16:2210 weakly suppresses
  5. L15:641 weakly suppresses
  6. L15:1345 weakly suppresses
  7. L16:1542 weakly suppresses
  8. L25:2323 weakly suppresses
  9. L27:1360 weakly suppresses
  10. L16:793 weakly excites
  11. L17:1322 weakly suppresses
  12. L20:332 weakly excites
  13. L17:963 weakly excites
  14. L16:2646 weakly excites
  15. L15:1603 weakly suppresses
  16. L16:1759 weakly excites
  17. L17:2897 weakly suppresses
  18. L27:269 weakly suppresses
  19. L17:1431 weakly suppresses
  20. L16:2304 weakly excites
  21. L19:2350 weakly suppresses
  22. L20:1238 weakly suppresses
  23. L16:2359 weakly excites
  24. L26:2595 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 -2.27 and -2.161 values between -1.84 and -1.731 values between -1.52 and -1.411 values between -1.41 and -1.303 values between -1.20 and -1.092 values between -1.09 and -0.983 values between -0.98 and -0.882 values between -0.88 and -0.778 values between -0.77 and -0.666 values between -0.66 and -0.5515 values between -0.55 and -0.4541 values between -0.45 and -0.34149 values between -0.34 and -0.23395 values between -0.23 and -0.121031 values between -0.12 and -0.021462 values between -0.02 and 0.09510 values between 0.09 and 0.20179 values between 0.20 and 0.3159 values between 0.31 and 0.4122 values between 0.41 and 0.527 values between 0.52 and 0.631 values between 0.63 and 0.733 values between 0.73 and 0.841 values between 0.84 and 0.951 values between 0.95 and 1.061 values between 1.16 and 1.271 values between 1.27 and 1.381 values between 1.49 and 1.591 values between 1.91 and 2.02offnegativepositiveoften

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. gratis
  2. olen
  3. .CreateInstance
  4. /public
  5. 建党
  6. iten
  7. ∋
  8. RLF
  9. inen
  10. реализаци

Suppresses

  1. eus
  2. 余
  3. 留
  4. 地段
  5. ерь
  6. tał
  7. ick
  8. 遗留
  9. �试
  10. queryParams