Neuron Explorer

Qwen3-0.6B

unclear: 'numerous' and 'several' in quantity talk

A neuron in layer 3 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron is hard to pin down. It lights up most on words meaning 'a lot of' or 'too many', like 'too numerous', but also on unrelated words such as 'spore', 'higher' and 'modes'. Turning it up made the model repeat itself or drift into odd topics, which does not point to a clear meaning.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    , or repel it. Perspicuity and conciseness are the most desirable qualities of a brief, but when the facts are material they cannot be too numerous when the argument is pertinent and weighty, it cannot be too extended.⏎Ec

  2. #2 peak

    protected as a species, but since the red wolf is in immediate danger of extinction, some steps are being made to reinstate its protection. Coyotes are too numerous to probably ever worry about them becoming extinct.⏎Gray wolves come in many colors.

  3. #3 peak

    reservoirs of the bacterium are herbivores, which harbour the bacteria in their bowels (with no consequences for them) and disseminate the “spore form” of the bacteria in the environment with their faeces.⏎Most cases

  4. #4 peak

    ; ground of argument.⏎2. (n.) The faculty or capacity of the human mind by which it is distinguished from the intelligence of the inferior animals; the higher as distinguished from the lower cognitive faculties, sense, imagination, and memory, and

  5. #5 peak

    coined the phrase, "Israelite People." Acting out of panic, Pharaoh wanted to do away with the Hebrews, when he declared they were becoming too numerous. (Exodus 1:9)⏎Even so, the distinction placed on

  6. #6 peak

    this technology. Its novel features, which include destruction of microorganisms at room temperature or lower, have made the technology commercially attractive. Enzymes and even spore forming bacteria can be inactivated by the application of pressure-thermal combinations, This

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 have a lot of people who are into this, but I also have to say that. I am very grateful for your time and effort. In the past, I have been able to provide a lot of people with this, and

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

    physiological fact irreconcilable with origin by descent with modification, would have been destructive of the theory and would have made the<|endoftext|>It is not good for the health of a tree to build up soil around its base.⏎In the Chicago suburbs and

  2. #2 peak

    . Elton, Jane Fairfax says,⏎“Excuse me, ma’am, but this is by no means my intention; I make no inquiry myself, and should be sorry to have any made by my friends. When I am quite determined as

  3. #3 peak

    is still a living force. It is more than a record of persecution. It is an arsenal of controversy, a storehouse of romance, as well as a source of edification.”⏎Fox’s Book of Martyrs is in the public domain

  4. #4 peak

    (v. t.) To arrange and present the reasons for or against; to examine or discuss by arguments; to debate or discuss; as, I reasoned the matter with my friend.⏎9. (v. t.) To support with reasons,

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. L2:1989 excites
  2. L1:853 excites
  3. L2:94 suppresses
  4. L2:1710 weakly suppresses
  5. L2:1 weakly suppresses
  6. L2:49 weakly suppresses
  7. L2:45 weakly excites
  8. L2:2783 weakly suppresses
  9. L0:1224 weakly suppresses
  10. L0:287 weakly suppresses
  11. L2:1321 weakly suppresses
  12. L1:146 weakly excites
  13. L1:1322 weakly excites
  14. L0:617 weakly suppresses
  15. L1:1057 weakly suppresses
  16. L2:843 weakly suppresses
  17. L1:2172 weakly suppresses
  18. L0:184 weakly suppresses
  19. L2:196 weakly excites
  20. L2:2485 weakly excites
  21. L2:41 weakly suppresses
  22. L2:499 weakly excites
  23. L0:1634 weakly suppresses
  24. L2:2405 weakly excites

Feeds later layers

  1. L7:956 suppresses
  2. L7:180 weakly suppresses
  3. L12:1666 weakly suppresses
  4. L10:481 weakly suppresses word before a noun naming people or a role
  5. L5:1149 weakly excites
  6. L14:1935 weakly suppresses
  7. L9:2630 weakly excites
  8. L7:1121 weakly suppresses
  9. L8:1654 weakly excites
  10. L10:863 weakly suppresses
  11. L26:2763 weakly excites
  12. L16:3010 weakly suppresses
  13. L13:2765 weakly suppresses
  14. L10:2905 weakly suppresses
  15. L16:2489 weakly suppresses
  16. L11:2820 weakly suppresses
  17. L9:1830 weakly suppresses
  18. L15:2662 weakly excites
  19. L8:2530 weakly suppresses
  20. L8:1857 weakly suppresses
  21. L5:17 weakly suppresses
  22. L19:1349 weakly suppresses
  23. L20:2283 weakly suppresses
  24. L8:2786 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 -3.55 and -3.411 values between -1.69 and -1.562 values between -1.56 and -1.421 values between -1.29 and -1.162 values between -1.16 and -1.032 values between -1.03 and -0.895 values between -0.89 and -0.767 values between -0.76 and -0.638 values between -0.63 and -0.5030 values between -0.50 and -0.3681 values between -0.36 and -0.23473 values between -0.23 and -0.101977 values between -0.10 and 0.041093 values between 0.04 and 0.17186 values between 0.17 and 0.3024 values between 0.30 and 0.436 values between 0.43 and 0.573 values between 0.70 and 0.834 values between 0.83 and 0.961 values between 0.96 and 1.101 values between 1.63 and 1.76offnegativepositiveoften

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. auce
  2. itate
  3. ationToken
  4. itar
  5. aterangepicker
  6. quần
  7. 鞭
  8. PageRoute
  9. Tops
  10. →finally

Suppresses

  1. 缺
  2. 珅
  3. scale
  4. /default
  5. 携
  6. %⏎
  7. sing
  8. ashi
  9. author
  10. 内分泌