Neuron Explorer

Qwen3-0.6B

"a" after sentence-opening "As"

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

This neuron lights up on the word "a" in phrases like "As a result" or "As a guide" at the start of a sentence. When it was turned up, the text drifted into unrelated talk about weather and temperatures, so the steering did not show a clear effect.

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

    , Norway, Sweden, and Finland. Researchers hope the study results will provide Native organizations and local governments across the Arctic with information to help make policy decisions. As a major step toward that goal, the team is hosting an international meeting with indigenous policy

  2. #2 peak

    personal lives might be put on the line.⏎It’s high time that all companies start intensifying their data protection measures to safeguard their data and hardware assets. As a guide, here are the five biggest IT security threats that you should prepare for as

  3. #3 peak

    China has also made agreements with Australia, Japan, Thailand, Russia, and Vietnam to allow for direct currency trade, instead of converting to the US Dollar. As a managed float, the Renminbi's value is determined by a basket of foreign

  4. #4 peak

    also impact the ways in which current and foreseeable future group individuals understand and interpret information that they receive. As a result, the relationship among communication and way of life is a complicated and romantic one

  5. #5 peak

    is still not officially considered to be medically underserved. When health professionals don't understand how to care for people with disabilities, it promotes bias in healthcare. As a result, patients with disabilities lose trust in medical care, which can put their lives

  6. #6 peak

    − tumors.”⏎Around 70% of breast cancers are ER-positive, meaning that the tumors have estrogen receptors and can grow in response to the hormone. As a result, ER-positive tumors can be treated in part with hormone-blocking therapies to restrict

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: "If you are a student, what are the implications of the internet?" This question is not just a personal one. It can be a question of "if you are a teacher" and "if you are a researcher." The internet is

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

    level n = 3 between x = 0 and x = L/ 6?⏎Answer to relevant QuestionsWhich has the lowest (ground-state) energy, an electron trapped in a one dimensional box of length 10-6 m or

  2. #2 peak

    1 we need to activate segments b and c.⏎This is the basic 8051 connected up to the 7 segment display, other circuitry such as crystal and capacitors and power is not shown here.⏎The display is a common

  3. #3 peak

    costs.⏎For the new study, published online in Critical Care Medicine, researchers sought to measure the impact of telemedicine consultations compared to other modes of treatment, such as telephone consultations or treatment without consultations.⏎The study included 320 seriously ill

  4. #4 peak

    Disc, Ashwathama thought of going to Krishna and getting the Sudarshana Chakra for himself by bargaining with the most powerful weapon he had – the Brahma-astra. He immediately left for Dwaraka and met Krishna.

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. L7:8 excites
  2. L7:135 excites 'of' after 'issue' in citations
  3. L6:1424 excites
  4. L6:60 excites
  5. L5:563 weakly excites
  6. L6:1639 weakly excites
  7. L6:228 weakly excites comma after a filler word
  8. L6:858 weakly excites
  9. L6:1479 weakly suppresses
  10. L6:2082 weakly excites
  11. L6:2594 weakly excites 'with' after a verb of action
  12. L2:2008 weakly excites
  13. L6:1641 weakly suppresses
  14. L6:512 weakly suppresses
  15. L2:896 weakly excites
  16. L4:16 weakly excites
  17. L6:662 weakly suppresses
  18. L6:1567 weakly excites colon introducing a summary point
  19. L6:2785 weakly excites
  20. L6:2178 weakly suppresses
  21. L6:1406 weakly excites
  22. L5:2939 weakly suppresses
  23. L3:2051 weakly excites
  24. L7:2456 weakly suppresses

Feeds later layers

  1. L9:459 suppresses
  2. L11:294 suppresses
  3. L20:94 suppresses
  4. L9:80 suppresses
  5. L12:652 suppresses
  6. L19:177 excites
  7. L16:1224 suppresses
  8. L10:459 suppresses
  9. L13:2594 excites
  10. L19:1446 suppresses
  11. L17:77 weakly excites
  12. L16:1548 weakly excites
  13. L16:316 weakly suppresses
  14. L11:1385 weakly excites
  15. L18:134 weakly excites
  16. L13:408 weakly suppresses
  17. L19:249 weakly excites
  18. L9:97 weakly excites
  19. L14:1690 weakly suppresses
  20. L15:35 weakly suppresses
  21. L21:225 weakly excites
  22. L16:786 weakly suppresses
  23. L9:283 weakly excites
  24. L14:1402 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 -5.50 and -5.331 values between -2.83 and -2.653 values between -2.65 and -2.472 values between -2.11 and -1.932 values between -1.93 and -1.761 values between -1.58 and -1.405 values between -1.40 and -1.2211 values between -1.22 and -1.046 values between -1.04 and -0.8617 values between -0.86 and -0.6933 values between -0.69 and -0.5156 values between -0.51 and -0.33266 values between -0.33 and -0.152176 values between -0.15 and 0.031013 values between 0.03 and 0.21204 values between 0.21 and 0.3957 values between 0.39 and 0.5732 values between 0.57 and 0.7412 values between 0.74 and 0.922 values between 0.92 and 1.103 values between 1.10 and 1.281 values between 1.28 and 1.464 values between 1.46 and 1.64offnegativepositiveoften

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. 意思
  2. altında
  3. ании
  4. spells
  5. okhttp
  6. 昆明
  7. orio
  8. odo
  9. 规定
  10. pytest

Suppresses

  1. 三级
  2. ERIC
  3. ersen
  4. tero
  5. atty
  6. .hasOwnProperty
  7. ers
  8. -than
  9. ctype
  10. bell