Neuron Explorer

Qwen3-0.6B

comma after a transition word

A neuron in layer 12 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 follows a linking word at the start of a sentence, such as "In turn,", "Instead,", "Furthermore," or "Similarly,". Turning it up made the model write repetitive, unrelated text, so the steering does not confirm the pattern.

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

    une Function, 2017, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5707683/#:~:text=It%20is%20a%20potent%

  2. #2 peak

    age of first-time mothers since the age of a woman’s first birth has a bearing on the total number of births she may have within her lifetime. In turn, this has an impact on family size and future increase in the overall population. In

  3. #3 peak

    diagnosed or undiagnosed depression often use drugs or alcohol as a form of self-medication. This then changes the brain structure and makes depression worse. Or, someone without depression starts using, and the changes in their brain cause depression.⏎In

  4. #4 peak

    capacity for this quality. Other experts disagree and argue that compassion isn’t the root cause of fatigue or stress. Instead, they say a general lack of tolerance<|endoftext|>Pack of 8 hanging succulents

  5. #5 peak

    be breakable even with today's computers, in which case the elliptic-curve algorithm will still provide the best security that today’s technology can offer. Alternatively, if the post-quantum algorithm turns out to be secure then it'll protect

  6. #6 peak

    building’s demand for electricity from the grid.⏎- Passive systems: The building design includes many passive features such as maximizing natural daylight and ventilation from the outdoors and, low-<|endoftext|>Five Questions about Long-Term Care⏎1. What is long

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: and I think, "and I have, and I have, and I will, and I will, and I will, and I will, and I will, and I will, and I will, and I will, and I will

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

    , a renowned American poet, has captivated readers with her profound exploration of nature and the human experience. Her poems often evoke a sense of wonder and contemplation, inviting readers to reflect on their own lives. In this argumentative essay, we

  2. #2 peak

    was decision-making within organizations and is best known for the theory of “bounded rationality” and the theory of “satisficing.” He applied psychology to statistics, operations, and decision-making scenarios to help define the process that people go through when

  3. #3 peak

    3. Symbolism: [Analyze the symbolic elements in the poem and their deeper meanings]⏎These devices enhance the reader’s understanding and emotional connection to the poem, making it a powerful piece of literature.⏎Impact on Readers⏎Mary Oliver’s poems

  4. #4 peak

    grow uncontrollably, forming a tumor. Early detection is crucial for successful treatment, and dogs have shown incredible abilities in detecting breast cancer at its early stages. These

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. L1:702 weakly excites
  2. L9:1952 weakly excites
  3. L6:1115 weakly suppresses
  4. L10:58 weakly excites
  5. L10:80 weakly excites
  6. L9:23 weakly suppresses
  7. L5:3066 weakly suppresses
  8. L11:223 weakly suppresses
  9. L9:103 weakly excites "as" or "since" giving a reason
  10. L4:244 weakly suppresses
  11. L5:2656 weakly excites
  12. L6:234 weakly excites
  13. L5:1511 weakly suppresses
  14. L10:39 weakly excites
  15. L6:2139 weakly suppresses
  16. L4:50 weakly excites
  17. L10:1523 weakly suppresses
  18. L8:311 weakly suppresses
  19. L1:2051 weakly suppresses
  20. L0:2593 weakly excites
  21. L9:618 weakly excites
  22. L10:1534 weakly suppresses
  23. L4:265 weakly excites
  24. L7:910 weakly suppresses

Feeds later layers

  1. L19:2741 excites
  2. L16:797 suppresses
  3. L17:1557 excites
  4. L17:2978 excites
  5. L13:2687 excites
  6. L15:1370 suppresses
  7. L16:2451 excites comma after an introductory generalization
  8. L21:639 weakly excites
  9. L14:1237 weakly suppresses
  10. L14:1158 weakly suppresses 'to' starting a purpose clause
  11. L18:391 weakly suppresses
  12. L20:660 weakly excites
  13. L15:3044 weakly excites
  14. L19:2085 weakly excites
  15. L21:721 weakly excites
  16. L19:433 weakly excites
  17. L21:191 weakly suppresses
  18. L18:956 weakly excites
  19. L16:528 weakly excites
  20. L24:1432 weakly excites
  21. L15:27 weakly excites
  22. L17:1495 weakly suppresses 'without' introducing a contrast
  23. L19:419 weakly excites
  24. L22:689 weakly excites comma closing an intro clause

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.

3 values between -2.14 and -2.032 values between -2.03 and -1.924 values between -1.92 and -1.811 values between -1.81 and -1.694 values between -1.47 and -1.363 values between -1.36 and -1.254 values between -1.25 and -1.133 values between -1.13 and -1.025 values between -1.02 and -0.919 values between -0.91 and -0.8015 values between -0.80 and -0.6919 values between -0.69 and -0.5719 values between -0.57 and -0.4651 values between -0.46 and -0.35107 values between -0.35 and -0.24258 values between -0.24 and -0.131020 values between -0.13 and -0.011792 values between -0.01 and 0.10368 values between 0.10 and 0.21125 values between 0.21 and 0.3225 values between 0.32 and 0.4415 values between 0.44 and 0.557 values between 0.55 and 0.6611 values between 0.66 and 0.775 values between 0.77 and 0.886 values between 0.88 and 0.996 values between 0.99 and 1.115 values between 1.11 and 1.221 values between 1.22 and 1.335 values between 1.33 and 1.442 values between 1.44 and 1.554 values between 1.55 and 1.671 values between 1.67 and 1.781 values between 2.00 and 2.112 values between 2.23 and 2.34offnegativepositiveoften

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. 可以获得
  2. 请联系
  3. operations
  4. -eyed
  5. уется
  6. みると
  7. Operational
  8. ationale
  9. обра�
  10. BaseEntity

Suppresses

  1. odor
  2. crollView
  3. 犁
  4. range
  5. ramid
  6. ulk
  7. licant
  8. //!
  9. uforia
  10. erty