Neuron Explorer

Qwen3-0.6B

and or comma closing a list

A neuron in layer 17 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

It fires on the "and" or comma that joins the last item of a list or the second half of a pair, such as "(1) ... and (2) ..." or "the stem stores food and supports...". It seems to track where a list of parallel parts is about to end. The steering outputs are mostly repetitive phrases like "on the other hand", which only loosely match this.

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

    ; (2) developing understanding of operations with rational numbers and working with expressions and linear equations; (3) solving problems involving scale drawings and informal geometric constructions, and working. K Core Curriculum - The Utah State Core Curriculum represents those standards of learning

  2. #2 peak

    ). Through iterative thematic analysis we developed an alliterative “7P” model spanning 2 domains: (1) Participants (Performers and Peers); and (2) Program (Partnerships, Practicalities, and Prevention).⏎Ex

  3. #3 peak

    roots. Each of these parts does a different job. The flower is where the plant reproduces; the leaf is where photosynthesis occurs; the stem stores food and supports the rest of the plant; and the roots absorb water and nutrients from the

  4. #4 peak

    the summed products of the following 2 factors: (i) the individuals beliefs that referents (such as family and friends) influence the behavior in question; and (ii) his or her motivation to comply with these referents. Research support

  5. #5 peak

    least developed countries as regards multilateral negotiations;⏎• Formulate a sound negotiation strategy;⏎• Analyze the significance of data required for a multilateral trade negotiation; and⏎• Assess various issues that may arise during negotiation

  6. #6 peak

    7).⏎Two expressions in the quotes above catch my attention: (a) “It may be difficult for you to envision…at one time…the grand narrative” and (b) “in contemporary society…the grand narrative has

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 am writing a story, which is a science fiction, fiction, fiction, fiction, fiction, fiction, fiction, fiction, fiction, fiction, fiction, fiction, fiction, fiction, fiction, fiction, fiction, fiction, fiction,

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

    and shape of a plant, and improve flowering and fruiting. There are four basic pruning techniques used for maintaining shrubs: heading back, renewal, rejuvenation, and shearing. For information on pruning specific shrubs, download the following list

  2. #2 peak

    3 primary big game species, pronghorn antelope (Antilocapra americana), mule deer (Odocoileus hemionus), and desert bighorn sheep (Ovis canadensis mexicana). Available

  3. #3 peak

    2019. The researchers analyzed life expectancy data for the 3 largest race/ethnicity groups by state: Hispanic, non-Hispanic Black, and non-Hispanic White Americans. They found that although mean life expectancy in

  4. #4 peak

    intellectual life more effectively.⏎Charlie Kurth, assistant professor in philosophy at Washington University in St Louis, Missouri, divides anxiety into three different categories: existential, social, and moral. The first, existential anxiety, is what we experience when we think

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. L16:2360 excites
  2. L15:941 excites
  3. L15:2440 excites word before 'and' or 'or' in pairs
  4. L16:2411 suppresses
  5. L13:557 excites
  6. L15:1839 excites
  7. L16:473 excites
  8. L15:3040 suppresses
  9. L15:132 excites
  10. L16:565 excites
  11. L15:1675 excites
  12. L16:2200 excites
  13. L15:2277 suppresses
  14. L16:1307 excites
  15. L14:449 excites
  16. L15:389 suppresses
  17. L15:2999 suppresses
  18. L16:1364 suppresses
  19. L15:1029 excites comma in three-item lists
  20. L15:2080 suppresses
  21. L15:137 suppresses
  22. L15:852 suppresses
  23. L16:2432 excites
  24. L16:1669 suppresses

Feeds later layers

  1. L19:845 strongly excites
  2. L19:1262 suppresses
  3. L18:2838 excites
  4. L21:344 excites
  5. L18:2918 excites
  6. L20:1083 excites
  7. L24:572 excites
  8. L20:699 excites
  9. L18:2862 excites
  10. L21:351 excites
  11. L18:2043 excites
  12. L20:82 suppresses
  13. L19:867 excites
  14. L18:957 excites
  15. L21:717 excites
  16. L20:756 excites
  17. L20:276 excites
  18. L19:164 suppresses
  19. L19:909 excites
  20. L20:328 excites
  21. L18:1404 excites
  22. L19:1092 excites
  23. L19:2229 excites
  24. L20:811 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 -9.22 and -8.811 values between -7.17 and -6.762 values between -6.76 and -6.351 values between -6.35 and -5.944 values between -5.94 and -5.532 values between -5.53 and -5.123 values between -5.12 and -4.713 values between -4.30 and -3.894 values between -3.89 and -3.483 values between -3.48 and -3.076 values between -3.07 and -2.6610 values between -2.66 and -2.255 values between -2.25 and -1.845 values between -1.84 and -1.4311 values between -1.43 and -1.0232 values between -1.02 and -0.61267 values between -0.61 and -0.202224 values between -0.20 and 0.211125 values between 0.21 and 0.6296 values between 0.62 and 1.0320 values between 1.03 and 1.4429 values between 1.44 and 1.8513 values between 1.85 and 2.2610 values between 2.26 and 2.679 values between 2.67 and 3.081 values between 3.08 and 3.495 values between 3.49 and 3.901 values between 3.90 and 4.317 values between 4.31 and 4.722 values between 4.72 and 5.131 values between 5.13 and 5.541 values between 5.54 and 5.951 values between 5.95 and 6.362 values between 6.36 and 6.771 values between 6.77 and 7.18offnegativepositiveoften

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. ;',
  2. ;',⏎
  3. ;,
  4. .',
  5. ],[
  6. ;",⏎
  7. ;'⏎
  8. .",
  9. !,
  10. .',⏎

Suppresses

  1. trx
  2. ducible
  3. .hh
  4. hua
  5. tones
  6. bara
  7. easy
  8. istar
  9. 飏
  10. andi