Neuron Explorer

Qwen3-0.6B

'of' after 'pair' or a landmark noun

A neuron in layer 0 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 responds to the word 'of' in phrases like 'pair of', 'tower of' or 'Gallery of'. It is quieter on 'of' after abstract nouns like 'implications of'. Turning it up did not produce a clear, consistent 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

    It was the first-fruits, and therefore, in the sanctification of it, the whole lump and building was sanctified" [POOLE].⏎the tower of Meah--This word is improperly considered, in our version, as the name

  2. #2 peak

    ston. For more information about this artist and her work, visit Inuit Gallery of Vancouver.<|endoftext|>1. The problem statement, all variables and given/known data

  3. #3 peak

    (at least to Physicists!) experiment, Galileo went to the top of the leaning tower of Pisa and dropped a wooden ball and a lead ball, both the same size

  4. #4 peak

    the eggs.⏎This is the second year of broadcasting a live feed from a bald eagle nest in the metro area. The DNR believes this is the same pair of eagles that have been using the nest for several years.⏎The camera site has

  5. #5 peak

    contractions with the arm in a pronated position for ten healthy and young male subjects with no overt sign of neuromuscular diseases. For a given pair of accelerometers, the spatial coherence of S-MMG is a measure of the similarity

  6. #6 peak

    frequently to help keep your eyes from feeling dry.⏎If you spend a considerable amount of time staring at your computer screen you might want to consider picking up a pair of blue light glasses. These special glasses help protect your eyes from LED light that can

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 most important thing is to be the most important. The most important thing is to be the most important. The most important thing is to be the most important. The most important thing is to be the most important. The most important

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

    of urology and examines both the pros and the cons of this emerging technology. The cost implications of this technology for healthcare facilities are considered, as well as the entrepreneurial opportunities that arise

  2. #2 peak

    territories of the global hemisphere, all aspects of human life were vastly affected. Especially, the educational sector brought in huge wonders. Students were shoved off the heavy burden of following the same boring⏎conventional educational methods. “Classroom flipping” was

  3. #3 peak

    fruit trees and grapevine genotypes may represent an important source of genes, with which in the future breeding programs can create varieties with high adaptability to adverse impacts of climate change. On the other hand, local collection of these genotypes will contribute

  4. #4 peak

    of the theoretical and empirical work within the field that has come to be known as Conservation Psychology . He notes that despite several decades of work focusing on the adverse impacts of human behavior, there appears to be a significant disconnect between theoretical models and empirical research

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

None: this is the first layer.

Feeds later layers

  1. L20:441 suppresses
  2. L22:866 weakly suppresses
  3. L23:1167 weakly suppresses
  4. L16:1518 weakly suppresses
  5. L26:1640 weakly suppresses
  6. L20:2217 weakly excites 'trying' after 'were' or 'are'
  7. L27:2344 weakly suppresses
  8. L18:2771 weakly excites
  9. L24:1957 weakly suppresses capital 'In' starting a word or name
  10. L17:127 weakly excites the word 'would' before a verb
  11. L26:383 weakly suppresses
  12. L8:2214 weakly suppresses
  13. L20:1630 weakly suppresses
  14. L25:2166 weakly suppresses
  15. L15:2108 weakly suppresses
  16. L12:1633 weakly suppresses
  17. L8:2212 weakly suppresses
  18. L19:2846 weakly suppresses
  19. L25:1553 weakly excites
  20. L6:2314 weakly excites 'by' after 'known' or 'referred to'
  21. L10:2016 weakly suppresses
  22. L24:2166 weakly excites
  23. L21:666 weakly suppresses
  24. L6:3016 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 -1.51 and -1.451 values between -1.38 and -1.311 values between -1.11 and -1.041 values between -1.04 and -0.981 values between -0.71 and -0.642 values between -0.57 and -0.513 values between -0.51 and -0.447 values between -0.44 and -0.3713 values between -0.37 and -0.3120 values between -0.31 and -0.2444 values between -0.24 and -0.17112 values between -0.17 and -0.10509 values between -0.10 and -0.042339 values between -0.04 and 0.03621 values between 0.03 and 0.10156 values between 0.10 and 0.1643 values between 0.16 and 0.2318 values between 0.23 and 0.302 values between 0.30 and 0.374 values between 0.37 and 0.433 values between 0.43 and 0.501 values between 0.50 and 0.571 values between 0.70 and 0.772 values between 0.77 and 0.832 values between 0.83 and 0.901 values between 1.10 and 1.17offnegativepositiveoften

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. objectId
  2. -coordinate
  3. borderBottom
  4. oined
  5. basePath
  6. userManager
  7. 中存在的
  8. 老字号
  9. mapped
  10. iked

Suppresses

  1. ment
  2. ments
  3. ner
  4. ыв
  5. TED
  6. นา
  7. lations
  8. VT
  9. ải
  10. 教师