Neuron Explorer

Qwen3-0.6B

comma after a 'that' or 'who' clause

A neuron in layer 6 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 a comma that comes right after a word like 'that' or 'who', where the sentence pauses to add an extra clause. Turning it up did not produce a clear, matching change in the written text.

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

    000 acres of pastures for cattle and sheep that, starting in the late 19th century, gave way to development of the

  2. #2 peak

    include:⏎- Limit or abstain from alcohol.⏎- Avoid overeating and, if overweight, implement healthy strategies to lose weight.⏎- Cut back on refined carbs

  3. #3 peak

    have considerably larger suppression costs and a much greater chance of escaping...⏎The ionospheric Alfvén resonator (IAR) is an area of space that, due to sharp Alfvén speed gradients, has the ability to trap electromagnetic plasma

  4. #4 peak

    the powerless, and the individual responsibility and autonomy of each person under God. In the New Testament, four different Greek words are used in reference to those people who, today, we call "Elders." A study of these four words -

  5. #5 peak

    Every data-driven trend comes with vulnerabilities that, if left unchecked and unprotected, can cause great damage.⏎In such cases, no

  6. #6 peak

    cards over three rounds, each new round begins after seven cards have been won. This is accomplished by translating cards in both languages. The winner is the player who, in the final round, translates all of the played cards error-free.⏎Novices

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: the number of the month of the number. I am not sure about the details of the plan. I need to think about something. I am not sure about the details of the plan. I need to think about something. I am not

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

    There are three possible levels at which this problem can be⏎tackled, and how it is used depends on the levels of the students.⏎As an exercise in confidence intervals it is relatively simple from⏎a calculational point of view; the

  2. #2 peak

    web-based systems.⏎So, despite spending millions of dollars on these ERP systems, organizations are still being relegated to the use of manual, paper-oriented processes and forms to fulfill various core

  3. #3 peak

    stamens 6, included, inserted at perianth base, anthers attached ± near middle; ovary ± sessile, style 1, entire or 3-branched⏎Fruit: capsule, loculicidal

  4. #4 peak

    The scientists could directly relate the lorikeets' deaths to inadequate artificial diets, and to the unsanitary conditions of trays and other containers from which they were being fed<|endoftext|>Published on March 1, 2013 by Carol⏎

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. L5:1217 weakly excites
  2. L1:1573 weakly suppresses
  3. L0:340 weakly suppresses
  4. L3:2118 weakly suppresses
  5. L2:2065 weakly excites
  6. L2:2388 weakly suppresses
  7. L2:2235 weakly excites
  8. L5:1506 weakly excites
  9. L3:305 weakly excites
  10. L0:2194 weakly suppresses
  11. L2:2897 weakly excites
  12. L2:156 weakly suppresses
  13. L4:1238 weakly excites
  14. L5:2657 weakly suppresses
  15. L5:713 weakly suppresses the spot after a comma in a number list
  16. L5:1858 weakly excites
  17. L0:807 weakly excites
  18. L3:1214 weakly excites
  19. L0:136 weakly suppresses space after a dash in number ranges
  20. L2:1870 weakly suppresses
  21. L1:1972 weakly suppresses
  22. L5:611 weakly excites
  23. L4:1539 weakly excites
  24. L4:631 weakly suppresses

Feeds later layers

  1. L9:459 suppresses
  2. L15:212 weakly suppresses
  3. L9:2267 weakly suppresses 'with' after exchange nouns
  4. L7:14 weakly suppresses
  5. L10:2279 weakly excites
  6. L22:201 weakly excites
  7. L22:761 weakly excites
  8. L17:769 weakly excites
  9. L12:2140 weakly excites 'the' after place prepositions
  10. L24:2543 weakly suppresses
  11. L23:1151 weakly suppresses
  12. L13:95 weakly suppresses
  13. L11:2199 weakly excites
  14. L18:1134 weakly suppresses
  15. L11:1392 weakly suppresses 'of' in fixed technical phrases
  16. L26:2134 weakly excites
  17. L14:126 weakly excites
  18. L21:2947 weakly suppresses
  19. L9:1562 weakly suppresses 'by' before a byline or agent
  20. L20:101 weakly suppresses
  21. L16:965 weakly excites
  22. L24:457 weakly excites
  23. L26:2424 weakly suppresses
  24. L15:622 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.

3 values between -1.20 and -1.031 values between -0.86 and -0.697 values between -0.69 and -0.5225 values between -0.52 and -0.35161 values between -0.35 and -0.181250 values between -0.18 and -0.011988 values between -0.01 and 0.16337 values between 0.16 and 0.3374 values between 0.33 and 0.5027 values between 0.50 and 0.6710 values between 0.67 and 0.847 values between 0.84 and 1.013 values between 1.01 and 1.196 values between 1.19 and 1.361 values between 1.36 and 1.532 values between 1.53 and 1.702 values between 1.87 and 2.042 values between 2.72 and 2.891 values between 3.40 and 3.571 values between 5.45 and 5.62offnegativepositiveoften

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. atra
  2. among
  3. 兮
  4. ập
  5. here
  6. /current
  7. -pop
  8. .Pop
  9. 说我
  10. periodically

Suppresses

  1. stro
  2. \Resource
  3. riding
  4. oice
  5. 力还是
  6. чная
  7. zte
  8. 裱
  9. 床位
  10. 很大的