Neuron Explorer

Qwen3-0.6B

citations right after author names

A neuron in layer 9 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 spot right after a source's name in a citation, like the gap before a year in '(Castaneda 1998)'. Turning it up made the model write more history-style or essay-style prose, but the effect was mixed.

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

    of symbols for multiple interpretations while allegory is the direct representation of concepts or ideas with a one on one relationship. “Sea of wisdom…” (Inferno 8, line 7) is symbolically used to represent a person who is

  2. #2 peak

    the movement known as the 26th July Movement was successful and the forces of Castro were successful in throwing up the government of Batista (Castaneda 1998, 80). Under the guidance of Castro, G

  3. #3 peak

    main accomplishments that characterized the American society. Through his characters, the novel The Great Gatsby by F. Scott Fitzgerald reveals the consequences of this dream on the population. The immorality of the characters of Jay Gatsby and Tom Buchanan is due

  4. #4 peak

    ⏎Che Guevara was born as Ernesto Guevara de la Serna in 1928 to a middle-class family (Castaneda 1998, 3). He was Argentinean by birth but was

  5. #5 peak

    contemptuous in that he tries to eat Bilbo a fellow hobbit only since he needs something to eat and raw fish are entirely his everyday meal (Chen 2012). This is an idea that is conveyed in the meme relating

  6. #6 peak

    organization have difficulty in reconciling their value paradigms with frameworks of action at least in the ethical domain. (Samir R Chattergee et al. 2000). Â The modern technological market ideology based societies need managers who

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: "This is a current in the mind. For example, if you check this list." Let's do this: For example, if you check this list, and also reply with "Hi, see the list" and then wait.

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

    setting of the story occur in Maycomb County, Alabama, throughout the 1930’s. # 3 Scout was telling the story about Boo Radley, she stated Radley was locked up in a basement as a teen for as

  2. #2 peak

    1-3.⏎In this study, we aimed to identify the dietary changes needed to achieve a nutritionally adequate diet (i.e.

  3. #3 peak

    ) were used to determine if there were significant differences in the extent of dominant manifestation of language learning strategies of the respondents when grouped according to the moderator variables. Results showed that grade VIII students moderately manifested the language learning strategies along the six categories. Both

  4. #4 peak

    and other small RNAs were extracted from specimens using TRIzol reagent. Qiagen's miScript System was used to synthesize cDNA and perform qPCR (Figure 1). Endogenous miRNAs are not polyadenylated

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. L8:2710 strongly suppresses
  2. L8:1223 suppresses
  3. L8:1819 excites
  4. L8:1242 excites
  5. L8:1292 suppresses
  6. L7:1143 excites
  7. L8:49 excites
  8. L3:448 excites
  9. L7:61 suppresses
  10. L8:1400 excites
  11. L8:1335 suppresses
  12. L7:2356 weakly suppresses
  13. L3:241 weakly excites
  14. L3:839 weakly suppresses
  15. L8:1980 weakly excites
  16. L1:1829 weakly excites
  17. L7:2883 weakly excites
  18. L8:2199 weakly excites
  19. L4:1883 weakly suppresses
  20. L8:341 weakly suppresses
  21. L5:268 weakly excites
  22. L7:63 weakly excites
  23. L3:1347 weakly excites
  24. L8:2809 weakly suppresses

Feeds later layers

  1. L11:2358 excites
  2. L10:2195 excites
  3. L10:2880 excites
  4. L15:995 suppresses
  5. L10:1042 excites
  6. L15:2714 excites
  7. L10:1672 excites
  8. L18:616 excites
  9. L11:2248 excites
  10. L10:2145 excites
  11. L20:250 weakly excites
  12. L18:504 weakly suppresses
  13. L15:883 weakly excites
  14. L13:1921 weakly excites
  15. L17:2505 weakly suppresses
  16. L17:2780 weakly suppresses
  17. L18:1287 weakly excites
  18. L12:244 weakly excites
  19. L15:1451 weakly excites
  20. L18:1756 weakly excites
  21. L12:1038 weakly excites
  22. L27:1821 weakly suppresses
  23. L21:2384 weakly suppresses
  24. L10:1112 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.

13 values between -0.89 and -0.7650 values between -0.76 and -0.64185 values between -0.64 and -0.51420 values between -0.51 and -0.39752 values between -0.39 and -0.27952 values between -0.27 and -0.14914 values between -0.14 and -0.02452 values between -0.02 and 0.1189 values between 0.11 and 0.2326 values between 0.23 and 0.3511 values between 0.35 and 0.486 values between 0.48 and 0.609 values between 0.60 and 0.735 values between 0.73 and 0.854 values between 0.85 and 0.972 values between 0.97 and 1.102 values between 1.10 and 1.221 values between 1.22 and 1.356 values between 1.35 and 1.471 values between 1.47 and 1.591 values between 1.59 and 1.721 values between 2.21 and 2.341 values between 2.34 and 2.461 values between 2.71 and 2.831 values between 3.21 and 3.332 values between 3.70 and 3.831 values between 3.95 and 4.07offnegativepositiveoften

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. rome
  2. echa
  3. mich
  4. beth
  5. imar
  6. rape
  7. icho
  8. iene
  9. doch
  10. InParameter

Suppresses

  1. 什么时候
  2. ��态
  3. 哪儿
  4. 什么地方
  5. 让孩子
  6. 什么意思
  7. forgotten
  8. 哪里
  9. [$_
  10. 什么