Neuron Explorer

Qwen3-0.6B

'therm' in science words

A neuron in layer 25 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 word piece 'therm' as in thermometer, thermodynamic or thermophilic. Turning it up gave a mix of topics, with a lot of 'art' and 'arts' words, which is only loosely related.

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

    the machine configuration file for assembly binding, see How the Runtime Locates Assemblies.⏎An application configuration file contains settings that are specific<|endoftext|>Mercury is used in thermometers, fluorescent light bulbs, and some electrical switches.⏎"Mercury" is

  2. #2 peak

    out more.⏎1. Maintain the Fridge Temperature⏎A recommended fridge temperature of 5 degrees C is advised. There are thermometers available at reasonable price which can help you determine the temperature. A report suggests

  3. #3 peak

    six breast pieces, each one might cook differently based on size. Don’t forget to clean your thermometer probe between readings to avoid cross-contamination.⏎Food or meat thermometers are fairly inexpensive, and will save you the worry of “is it done

  4. #4 peak

    ; the indirect techniques only calculate it from other indirect indicators.⏎Image Credit: Soonthorn Wongsaita / Shutterstock⏎Many devices for LOD measurement use the thermogravimetric principle. This uses the total loss of weight incurred by the

  5. #5 peak

    use of epoxy-based resins and may be classified as the thermosetting epoxide polymer. Exteriors that receive this coating are resistant to

  6. #6 peak

    activation energy to break covalent bonds and⏎denature/destroy molecular structure. Proteins are particularly sensitive.⏎There are exceptions, as in the case of thermophillic bacteria. There are⏎cases or heat shock-proteins or "

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: being in this lесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесесес

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

    (28.2 cm.), W: 8 7/8 in. (22.5 cm.)|⏎|Location||Alte Pinakothek, Bayerische Staatsgemä

  2. #2 peak

    Museum of Alberta, Musée national des arts et traditions populaires, Département de l'organisation des systèmes d'information,<|endoftext|>A Very Short Introduction⏎Liberalism is one

  3. #3 peak

    the historical development of the West.⏎The first link is ideological. Nazism emerged in the socio-political constellation of German nationalism, which was crisscrossed by currents well represented

  4. #4 peak

    language Kasak “nomad”, “free man”, the term is first mentioned in a Ruthenian document dating from 1395.⏎In socio-political terms, from the 15th century to the present day, the

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. L24:105 weakly suppresses
  2. L21:463 weakly excites
  3. L4:1572 weakly excites
  4. L20:959 weakly excites
  5. L0:917 weakly suppresses
  6. L6:2695 weakly excites
  7. L8:1699 weakly suppresses
  8. L19:1351 weakly suppresses
  9. L22:164 weakly suppresses
  10. L10:2172 weakly suppresses
  11. L3:2050 weakly excites
  12. L24:2872 weakly suppresses
  13. L13:481 weakly suppresses 'because' introducing a reason
  14. L15:1237 weakly suppresses
  15. L5:1301 weakly excites
  16. L20:390 weakly suppresses
  17. L7:2308 weakly suppresses comma inside stacked adjectives
  18. L23:1427 weakly suppresses
  19. L14:1304 weakly suppresses
  20. L7:49 weakly suppresses
  21. L0:801 weakly suppresses 'long' in 'how long' questions
  22. L8:2688 weakly excites
  23. L8:1487 weakly excites
  24. L7:1221 weakly suppresses

Feeds later layers

  1. L27:787 suppresses
  2. L26:984 weakly suppresses
  3. L27:1599 weakly suppresses
  4. L27:1233 weakly suppresses
  5. L27:2675 weakly excites
  6. L27:178 weakly excites
  7. L26:2948 weakly suppresses
  8. L27:1012 weakly suppresses
  9. L26:2594 weakly suppresses
  10. L27:2179 weakly excites
  11. L27:2423 weakly excites
  12. L27:618 weakly excites
  13. L27:572 weakly excites
  14. L27:1119 weakly excites
  15. L27:1055 weakly suppresses
  16. L26:2197 weakly suppresses
  17. L27:494 weakly excites
  18. L27:2278 weakly suppresses
  19. L27:2305 weakly excites
  20. L27:2044 weakly excites
  21. L26:1836 weakly excites
  22. L26:1626 weakly excites the word 'I' as subject
  23. L27:1822 weakly excites
  24. L26:2186 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 -27.73 and -26.621 values between -26.62 and -25.501 values between -21.04 and -19.931 values between -17.70 and -16.581 values between -16.58 and -15.472 values between -14.35 and -13.241 values between -13.24 and -12.124 values between -12.12 and -11.012 values between -11.01 and -9.892 values between -9.89 and -8.783 values between -8.78 and -7.664 values between -7.66 and -6.552 values between -6.55 and -5.434 values between -5.43 and -4.318 values between -4.31 and -3.209 values between -3.20 and -2.0824 values between -2.08 and -0.973030 values between -0.97 and 0.15769 values between 0.15 and 1.2619 values between 1.26 and 2.387 values between 2.38 and 3.493 values between 3.49 and 4.614 values between 4.61 and 5.721 values between 5.72 and 6.841 values between 6.84 and 7.951 values between 9.07 and 10.182 values between 10.18 and 11.301 values between 15.76 and 16.88offnegativepositiveoften

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. art
  2. artin
  3. bow
  4. artisan
  5. indicator
  6. art
  7. иж
  8. Arn
  9. asted
  10. vation

Suppresses

  1. emat
  2. èmes
  3. erah
  4. ème
  5. etu
  6. eta
  7. eti
  8. eton
  9. eto
  10. ém