Neuron Explorer

Qwen3-0.6B

"also" before a verb

A neuron in layer 15 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 "also" when it sits between a subject or helper verb and the main verb, as in "who also claimed" or "will also sink". Turned up, the model started repeating words like "same" and "sunny", which loosely fits the idea of adding or repeating.

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

    ." In 1985, the Journal of Dental Research published a report by dentists Murry Vimy and F.L. Lorscheider, who also claimed that mercury vapor is released from amalgam fillings. These researchers further stated

  2. #2 peak

    Prime Minister in Britain, and initiated her radical reforms and a long period of growth. A year later, Ronald Reagan was elected President of the United States, and also embraced free market policies. All three of these leaders professed inspiration from the work

  3. #3 peak

    performance careers, and up to 59 percent of foals—particularly in foals raised in confinement. Colonic ulcers are only slightly less common, also affecting actively used horses.⏎Unfortunately, ulcers

  4. #4 peak

    Ns armed with 800 kg armor piercing bombs, in four sections. 40 B5Ns armed with Type 91 torpedoes, also in four sections).⏎- 2nd Group – targets: Ford Island and Wheeler

  5. #5 peak

    on the other hand, encompass the company's financial obligations and debts. These can include loans, accounts payable, and accrued expenses. Similar to assets, liabilities can also be classified as current or long-term, depending on their repayment timeline.⏎Shareholders

  6. #6 peak

    and he has talked and written extensively on these subjects. He is a member of the Industrial Workers of the World, following his father William Chomsky who was also a member.⏎Notes[change | change source]⏎- "Chomsky is

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 have been told that I have a problem with my body and have developed a problem with my mind and have developed a problem with my body and have developed a problem with my mind and have developed a problem with my body and have developed a

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

    olic blood pressure (the top number) decreased by 8 to 9 mmHg, on average. Diastolic blood pressure (the bottom number) decreased by 6 to 7 mmHg, on average. The higher the

  2. #2 peak

    of<|endoftext|>In the past 2 years, e-cigarette television advertising in the US directed at youth has increased two-fold, while ads directed at young adults have increased three-fold, according to a study by researchers at RTI International and the Florida

  3. #3 peak

    metry. In 65% of all individuals the left planum temporale appears to be more developed, while the right planum temporale is more developed in only 10%. In some people’s brains, the planum temporale

  4. #4 peak

    Nothing could be further from the truth. In fact Alcock was trained and flew in the RNAS (Royal Naval Air Service) and Brown was trained and flew in the RFC (Royal Flying Corps). And, both took their flying duties very seriously

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. L14:1425 excites
  2. L14:2983 excites
  3. L12:193 suppresses
  4. L5:979 weakly excites
  5. L12:2794 weakly excites
  6. L13:2317 weakly excites
  7. L14:630 weakly excites
  8. L3:554 weakly excites
  9. L13:772 weakly suppresses
  10. L12:135 weakly suppresses
  11. L2:6 weakly excites
  12. L14:563 weakly excites
  13. L14:505 weakly excites
  14. L12:723 weakly suppresses
  15. L14:113 weakly suppresses
  16. L1:2255 weakly suppresses
  17. L13:637 weakly suppresses opening parenthesis after a term
  18. L9:22 weakly excites
  19. L10:1830 weakly excites
  20. L12:132 weakly suppresses
  21. L12:62 weakly suppresses
  22. L11:798 weakly excites
  23. L11:518 weakly suppresses
  24. L3:429 weakly suppresses

Feeds later layers

  1. L17:1606 excites
  2. L16:2179 excites
  3. L16:2567 excites
  4. L17:1497 suppresses
  5. L17:1586 suppresses
  6. L19:1462 excites
  7. L18:812 suppresses
  8. L17:429 suppresses
  9. L19:775 excites
  10. L21:560 excites
  11. L16:1355 suppresses
  12. L16:1128 excites
  13. L18:556 excites
  14. L19:1530 weakly suppresses
  15. L17:861 weakly excites
  16. L16:1525 weakly excites
  17. L17:2419 weakly suppresses
  18. L16:1610 weakly suppresses
  19. L16:2439 weakly excites
  20. L16:2075 weakly suppresses
  21. L17:1160 weakly excites unclear: inside numbers and names
  22. L16:1151 weakly excites
  23. L18:154 weakly excites
  24. L19:2300 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 -2.36 and -2.171 values between -1.42 and -1.231 values between -1.23 and -1.041 values between -1.04 and -0.852 values between -0.85 and -0.6623 values between -0.66 and -0.48126 values between -0.48 and -0.29807 values between -0.29 and -0.102111 values between -0.10 and 0.09690 values between 0.09 and 0.28114 values between 0.28 and 0.4710 values between 0.47 and 0.665 values between 0.66 and 0.844 values between 0.84 and 1.035 values between 1.03 and 1.221 values between 1.41 and 1.601 values between 1.60 and 1.791 values between 1.79 and 1.971 values between 2.73 and 2.921 values between 4.05 and 4.231 values between 4.23 and 4.421 values between 4.99 and 5.18offnegativepositiveoften

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. 同样的
  2. same
  3. 同一个
  4. 同样
  5. same
  6. Same
  7. 一样
  8. 一样的
  9. 一樣
  10. }}</

Suppresses

  1. askan
  2. 蛱
  3. iquid
  4. xp
  5. år
  6. 즘
  7. 小时前
  8. itu
  9. практи
  10. obs