Neuron Explorer

Qwen3-0.6B

'the' after a verb or preposition

A neuron in layer 7 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 lights up on the word 'the' when it comes right after a verb or a preposition, such as 'constructed the' or 'shield themselves from the'. Turning it up made the writing wander into odd, loosely connected questions.

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

    ars was located on a stretch of the river called "Deer Park Bend". In 1933, the U.S. Army Corps of Engineers constructed the "Glasscock Cutoff", which removed Briers from the contiguous flow of the

  2. #2 peak

    of flowers, and a number of medicinal plants is invariably present.⏎Throughout the newer parts of the country it is seen that the conditions governing the maintenance and use of the vegetable garden are somewhat different, and, while a number of vegetable crops may be

  3. #3 peak

    to experience painful, damaging sunburns, a new study shows.⏎The study, the first to look at how people in the U.S. shield themselves from the sun and how well those different strategies work, analyzed information on more than 3

  4. #4 peak

    of units⏎The cgs electromagnetic system of units⏎The cgs electric units were much too small for practical use in engineering, which lead to the creation of the International System of units (not be

  5. #5 peak

    and adults with disabilities using the latest assistive technology.⏎Communicating with Technology⏎Find out about synthetic speech systems, speech to text and screen readers. Great for the visually or hearing impaired student.⏎A Language Arts Lesson Plan on Summarizing Text

  6. #6 peak

    to conduct current on applying a potential, across it. Flame rods are normally used for detecting flames from gas. Oil generated flames are not eligible for the application of the flame rods. The flame rods comes with the following advantages.⏎• It provides a

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: of a new, powerful and effective way of using technology to improve the quality of life of many people, of which one has introduced, in the past, a number of health problems in the management of the quality of food of food of a

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

    that school resources, student academic ability, and students learning in their mother tongue are some factors for consideration. Copyright © 2014 Society for the Scientific Study of Reading (SSSR).⏎|Publication status||Published - Jul

  2. #2 peak

    Annual Report noted that children younger than 15 accounted for 26 percent of those injuries.⏎Prevent Blindness America announced it supports bans on the importation, sale and use of all fireworks and sparklers by non-licensed operators,

  3. #3 peak

    iveness of a tiered intervention model in Chinese: A growth model of reading fluency. Paper presented at the Twenty-First Annual Meeting: Society for the Scientific Study of Reading, Santa Fe, New Maxico.⏎- Response to intervention<|endoftext|>Much

  4. #4 peak

    Consulting Meteorologist (CCM)", many of whom have expertise in the applied meteorology discipline of atmospheric dispersion modeling⏎Atmospheric dispersion modeling is the mathematical simulation of

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. L4:49 suppresses
  2. L4:73 suppresses
  3. L2:1390 suppresses
  4. L2:205 weakly excites
  5. L4:740 weakly suppresses
  6. L1:911 weakly excites
  7. L1:1001 weakly excites
  8. L5:1092 weakly suppresses
  9. L4:163 weakly suppresses
  10. L5:153 weakly suppresses
  11. L3:344 weakly excites
  12. L2:1790 weakly suppresses
  13. L2:2202 weakly excites
  14. L6:640 weakly suppresses "the" introducing a noun phrase after a name or preposition
  15. L6:450 weakly excites
  16. L0:1778 weakly suppresses
  17. L6:2092 weakly suppresses
  18. L3:1370 weakly suppresses
  19. L1:1693 weakly suppresses
  20. L6:1679 weakly excites 'and' joining two actions in instructions
  21. L6:1684 weakly suppresses
  22. L2:690 weakly excites
  23. L6:18 weakly excites
  24. L4:417 weakly suppresses

Feeds later layers

  1. L9:1155 excites
  2. L13:1234 excites
  3. L10:1477 excites
  4. L9:2601 excites 'the' after verbs in advice lists
  5. L12:2287 suppresses
  6. L12:1364 excites
  7. L10:1162 suppresses
  8. L18:1547 excites
  9. L11:875 excites
  10. L17:1592 suppresses
  11. L9:226 suppresses 'to' starting an infinitive
  12. L16:2062 weakly excites "other" and contrast words
  13. L15:1848 weakly excites
  14. L16:2248 weakly suppresses
  15. L10:3028 weakly excites
  16. L14:727 weakly suppresses
  17. L12:1990 weakly excites
  18. L19:379 weakly excites
  19. L19:2893 weakly suppresses
  20. L17:806 weakly suppresses
  21. L11:1245 weakly suppresses
  22. L10:1704 weakly excites
  23. L15:1176 weakly suppresses
  24. L14:1402 weakly excites

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 -2.28 and -2.191 values between -2.09 and -2.001 values between -2.00 and -1.902 values between -1.71 and -1.612 values between -1.42 and -1.322 values between -1.32 and -1.231 values between -1.23 and -1.131 values between -1.04 and -0.941 values between -0.94 and -0.843 values between -0.84 and -0.752 values between -0.75 and -0.655 values between -0.65 and -0.569 values between -0.56 and -0.4633 values between -0.46 and -0.36115 values between -0.36 and -0.27265 values between -0.27 and -0.17694 values between -0.17 and -0.081318 values between -0.08 and 0.02877 values between 0.02 and 0.12374 values between 0.12 and 0.21122 values between 0.21 and 0.3141 values between 0.31 and 0.4012 values between 0.40 and 0.506 values between 0.50 and 0.606 values between 0.60 and 0.693 values between 0.69 and 0.791 values between 0.79 and 0.881 values between 0.88 and 0.983 values between 0.98 and 1.084 values between 1.08 and 1.171 values between 1.17 and 1.271 values between 1.46 and 1.56offnegativepositiveoften

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. qui
  2. 高出
  3. via
  4. 给您
  5. iosis
  6. ays
  7. (IS
  8. IVES
  9. 经
  10. ients

Suppresses

  1. shed
  2. $d
  3. ỡ
  4. MERCHANTABILITY
  5. òn
  6. 大宗商品
  7. ielding
  8. nombreux
  9. ocy
  10. shr