Neuron Explorer

Qwen3-0.6B

sentence-ending periods before new sentences

A neuron in layer 11 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 full stop or line break that ends a sentence, right before a new one starts, especially after a name or place. When it was turned up, the text drifted into odd, unrelated topics, so it does not clearly steer toward one theme. When turned down, it produced repetitive text.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    Year 2.⏎Year 1 follow Curriculum Music⏎Year 2 follow ‘Listen 2 Me

  2. #2 peak

    Naval Cemetery. You can still visit a memorial erected in their honor in Hagatna today.⏎

  3. #3 peak

    Research Report. Book 4.<|endoftext|>Sardar Sarovar Dam, a project for

  4. #4 peak

    1944.⏎Others are Kolmaneskuppe, Namibia, a former diamond mining town whose

  5. #5 peak

    state of Utah. As of the 2010 census, the population was 23

  6. #6 peak

    25) and 0.72 (95% CI = 0.52–0.86; n = 27), respectively. Results indicated that body size, standardized by sex, was the most important cov

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: how to use a calculator in a real situation, how to use my brain, how to use my time, how to keep moving, how to do this, how to be intentional, how to learn, how to solve a common mistake,

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

    Llanrhydd. Llannerch contained the parishes of Llanfair D.C., Llanelidan and Derwen,<|endoftext|>Ulcers are a common scourge in horses. Gastric ulcers affect up to

  2. #2 peak

    center regularly updates its Web site with new resources about different content areas and offers poster sessions at the January and July residencies. LibGuides: Commerce Research Guide: Where can find journal ...⏎Functioning as a research and publishing network, ScienceOpen

  3. #3 peak

    for about 23 percent of the oil and 8 percent of total natural gas produced in the United States (approximately 82 percent of the oil supply comes from deepwater facilities), and the Gulf Coast region is home to almost half of

  4. #4 peak

    ography Visual Data: Photography, Film, and Video Using Documents as Data Qualitative Online Research: Using the Internet Observation and Mediated Data: An Overview PART SIX: FROM TEXT

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. L7:911 suppresses
  2. L7:63 excites
  3. L7:65 excites
  4. L10:762 suppresses
  5. L9:1308 weakly suppresses
  6. L5:654 weakly excites
  7. L9:1033 weakly suppresses unclear: end-of-text markers and word fragments
  8. L7:1130 weakly excites
  9. L10:258 weakly excites
  10. L10:1015 weakly excites
  11. L3:1703 weakly suppresses
  12. L10:284 weakly suppresses
  13. L8:4 weakly excites
  14. L3:58 weakly excites
  15. L8:183 weakly suppresses
  16. L5:821 weakly excites
  17. L7:424 weakly suppresses
  18. L7:12 weakly suppresses
  19. L6:898 weakly excites
  20. L5:64 weakly excites
  21. L10:959 weakly suppresses
  22. L0:1380 weakly suppresses
  23. L2:1995 weakly excites
  24. L3:719 weakly suppresses digits inside numbers and statistics

Feeds later layers

  1. L13:617 excites
  2. L14:1620 excites
  3. L12:2726 excites
  4. L13:1242 excites
  5. L13:496 suppresses
  6. L13:2091 excites
  7. L13:1589 suppresses
  8. L14:1473 excites
  9. L12:190 excites
  10. L13:808 suppresses
  11. L13:1468 excites
  12. L14:912 suppresses
  13. L13:2156 excites
  14. L14:1533 excites
  15. L13:429 excites
  16. L13:619 excites
  17. L13:997 excites
  18. L14:398 excites
  19. L15:820 suppresses
  20. L13:2186 excites
  21. L12:546 excites
  22. L12:28 suppresses
  23. L15:2660 excites
  24. L15:150 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 -2.40 and -2.291 values between -1.73 and -1.621 values between -1.51 and -1.402 values between -1.40 and -1.283 values between -1.17 and -1.067 values between -1.06 and -0.9510 values between -0.95 and -0.847 values between -0.84 and -0.7314 values between -0.73 and -0.6118 values between -0.61 and -0.5031 values between -0.50 and -0.3978 values between -0.39 and -0.28200 values between -0.28 and -0.17666 values between -0.17 and -0.061834 values between -0.06 and 0.06713 values between 0.06 and 0.17215 values between 0.17 and 0.2852 values between 0.28 and 0.3929 values between 0.39 and 0.509 values between 0.50 and 0.614 values between 0.61 and 0.733 values between 0.73 and 0.842 values between 0.95 and 1.061 values between 1.06 and 1.173 values between 1.17 and 1.282 values between 1.40 and 1.511 values between 1.73 and 1.841 values between 1.95 and 2.06offnegativepositiveoften

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. antu
  2. migrationBuilder
  3. (ierr
  4. kat
  5. rom
  6. 延
  7. uai
  8. 的那种
  9. 一面
  10. Hait

Suppresses

  1. config
  2. ��
  3. 来这里
  4. calc
  5. eldon
  6. icles
  7. ...")⏎
  8. andles
  9. anatomy
  10. bullet