Neuron Explorer

Qwen3-0.6B

'of' after a quality or importance adjective

A neuron in layer 19 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

The neuron wakes up on the word 'of' when it follows a description of how important, big, or good something is, as in 'of considerable importance' or 'of little interest'. It tends to go quiet on 'of' in fixed phrases like 'out of place'. Turning it up made the model drift into unrelated text, so the link to the word itself is the main clue.

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

    primary process, few stop to ask a very important question: why do we have primaries in the first place? In the history of American democracy, political primaries are of relatively recent origin, so when and why did we adopt them?⏎Before answering these

  2. #2 peak

    male population and on the right of the female population, thereby indicating the sexes. Pyramids are of two types: One, based on absolute numbers of population and the second, on

  3. #3 peak

    blackens the leaves; pull up the plants and discard them.<|endoftext|>Introduction and general survey⏎The economic importance of power amplifiers⏎Audio power amplifiers are of considerable economic importance. They are built in their hundreds of thousands every year, and

  4. #4 peak

    , complex ventures along military lines. Between 1660 and 1727, 87% of the colonial governors were British army officers of the rank of colonel or higher, with an average of ten years' military

  5. #5 peak

    passage and are rather interesting. The remaining three Column Cave, Keedy Cave, and Rohrersville No. 5 Cave are small and of little interest. All but the latter occur within 1000 feet of

  6. #6 peak

    ened grapes of the main harvest, typically picked in September, and are usually made in a light style. The term ‘Kabinette’ originally implied a wine of superior quality, set aside for later sale.⏎The term originated with the cister

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: how to use a pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern pattern

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

    scene gives one “a reason to be there” in situations where even senior researchers might struggle with a feeling of being “out of place.” It also helps to fight boredom in the field, serves as an opportunity

  2. #2 peak

    powerless to improve their own situation in life. This means that they become unwilling to take action to improve a situation because they have already decided that it will not make much difference. An example of this would be the individual who is addicted to alcohol.

  3. #3 peak

    projects in the state’s current coastal Master Plan, which incorporated worst-case scenarios for relative sea-level rise calculated two years ago— which the new figures now make out-of-date. (The state’s estimates of sea-level rise and subsidence are listed

  4. #4 peak

    acknowledge our discomfort in dealing with the subject.⏎You didn’t indicate whether your 8-year-old was a boy or a girl, but it wouldn’t really make much difference anyway. The best approach is an honest one that doesn’t try to offer

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. L18:2711 suppresses
  2. L18:2998 excites
  3. L17:2436 suppresses
  4. L5:75 weakly excites
  5. L4:1523 weakly suppresses
  6. L18:164 weakly suppresses
  7. L18:828 weakly suppresses
  8. L16:432 weakly excites period after month or initial
  9. L7:2705 weakly suppresses
  10. L10:331 weakly suppresses
  11. L17:3009 weakly excites
  12. L9:2196 weakly excites
  13. L2:2985 weakly suppresses
  14. L16:2790 weakly excites
  15. L7:135 weakly excites 'of' after 'issue' in citations
  16. L18:2087 weakly suppresses "with" after a group noun
  17. L14:1273 weakly excites
  18. L3:602 weakly excites 'the' after 'of' in noun phrases
  19. L2:349 weakly excites
  20. L18:1278 weakly excites 'a' or 'most' before a quantity
  21. L16:2120 weakly suppresses
  22. L8:1105 weakly excites
  23. L18:2476 weakly suppresses
  24. L4:4 weakly suppresses

Feeds later layers

  1. L20:2393 suppresses
  2. L27:488 suppresses
  3. L23:142 excites
  4. L20:1284 suppresses
  5. L22:404 weakly suppresses
  6. L26:528 weakly suppresses
  7. L22:240 weakly excites
  8. L24:200 weakly excites
  9. L21:274 weakly excites
  10. L21:323 weakly suppresses
  11. L24:1079 weakly excites
  12. L21:599 weakly suppresses
  13. L25:99 weakly suppresses
  14. L22:837 weakly suppresses
  15. L21:975 weakly suppresses
  16. L20:526 weakly suppresses
  17. L23:473 weakly excites
  18. L23:1135 weakly excites
  19. L23:360 weakly suppresses
  20. L22:2587 weakly suppresses
  21. L21:68 weakly excites
  22. L27:592 weakly suppresses
  23. L20:731 weakly excites
  24. L20:2109 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 -17.80 and -17.161 values between -10.74 and -10.091 values between -10.09 and -9.451 values between -9.45 and -8.811 values between -8.17 and -7.532 values between -6.89 and -6.242 values between -6.24 and -5.601 values between -5.60 and -4.961 values between -4.96 and -4.324 values between -4.32 and -3.6811 values between -3.68 and -3.0319 values between -3.03 and -2.3920 values between -2.39 and -1.7546 values between -1.75 and -1.11156 values between -1.11 and -0.472851 values between -0.47 and 0.18712 values between 0.18 and 0.8230 values between 0.82 and 1.4613 values between 1.46 and 2.1014 values between 2.10 and 2.747 values between 2.74 and 3.395 values between 3.39 and 4.033 values between 4.03 and 4.673 values between 4.67 and 5.311 values between 5.31 and 5.951 values between 5.95 and 6.601 values between 7.24 and 7.88offnegativepositiveoften

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. ingen
  2. riger
  3. unist
  4. ing
  5. asser
  6. рав
  7. ylvania
  8. 植入
  9. 自负
  10. 报记者

Suppresses

  1. 意义
  2. 意義
  3. reason
  4. 理由
  5. смысл
  6. detail
  7. 道理
  8. duration
  9. 眼光
  10. chance