Neuron Explorer

Qwen3-0.6B

"that" starting a relative clause

A neuron in layer 6 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 "that" when it follows a noun and begins a describing clause, as in "materials that review grammar" or "links that support learning". Turning it up or down did not give writing that clearly reflects this pattern.

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's definitions, example sentences, related words, idioms and quotations.⏎More words from Urdu related to Communized⏎View an extensive list of words below that are related to the meanings of the word Communized meanings in Urdu in Urdu.⏎

  2. #2 peak

    your comprehension.⏎Grammar topics fill the second content area of the exam and consist of written excerpts and questions on grammar, vocabulary and syntax. Many study guides are available that review grammar topics and include exercises to help refine your grammar skills. In addition to

  3. #3 peak

    Freewrite for 15 minutes in your journal about what it means to be an educated person. To get started, think about someone you know or admire that you consider to be an example of an educated person.⏎2. Working with a

  4. #4 peak

    the vaccine industry. There are some different vaccine development methods on the horizon, however, that will use bacteria that have been genetically engineered to make flu vaccines. But until that

  5. #5 peak

    examples in lower grades or with minor faults sell for a surprising percentage of Scott catalog value.⏎Determine what you can afford and buy the best example you can find that

  6. #6 peak

    even color changes to their Ozobot to reflect the character’s feelings and actions.⏎Edpuzzle gives you the opportunity to easily create interactive video lessons for your students that you can integrate right into your Learning Management System (LMS). Join us as

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: what is the value of x, x? March 14th, 2023, 23:55:23 AM, 23:55, 23:55, 2

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

    resolution of disputes between disputants within reasonable time with mo additional cost.⏎(3) WHAT HAPPENS IN A MEDIATION -Mediation is a structured process and Comprises of various stages, which are as under :- 1. Introduction :-

  2. #2 peak

    Bible as a textbook—and worshiping God according to the revelation within its pages were key components<|endoftext|>Hey There, welcome and warm greetings. Below an interesting article and at the end my own speculation and not endorsed by the source providers, purely my

  3. #3 peak

    facilitate habits and, for some, sustained compulsivity and addiction. Like other complex illnesses and disorders, addiction is multifactorial, resulting from a combination of genetic, social

  4. #4 peak

    ite) exoskeletons, giving them structural rigidity. These colonial “hard corals” form elaborate finger-shaped, branching, or moundshaped structures and can create masses of limestone that stretch

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. L1:2683 excites
  2. L5:322 weakly suppresses
  3. L2:483 weakly suppresses
  4. L5:225 weakly suppresses
  5. L2:66 weakly excites
  6. L1:1031 weakly suppresses
  7. L2:1710 weakly suppresses
  8. L1:2325 weakly suppresses
  9. L4:2517 weakly suppresses
  10. L3:59 weakly excites
  11. L4:2756 weakly suppresses
  12. L2:1162 weakly suppresses
  13. L5:653 weakly suppresses
  14. L5:1851 weakly suppresses
  15. L5:110 weakly suppresses
  16. L3:1744 weakly excites
  17. L1:2443 weakly suppresses
  18. L2:70 weakly excites
  19. L3:1905 weakly excites
  20. L5:1958 weakly excites
  21. L1:1833 weakly excites
  22. L3:1030 weakly excites
  23. L4:547 weakly suppresses
  24. L5:129 weakly excites

Feeds later layers

  1. L9:661 excites
  2. L7:1996 excites
  3. L17:1361 excites 'that' after verbs like recommend
  4. L11:1278 excites
  5. L7:2390 excites
  6. L8:94 excites
  7. L17:1008 excites
  8. L15:1902 excites
  9. L10:1256 weakly excites
  10. L7:2503 weakly excites
  11. L19:1185 weakly excites
  12. L9:865 weakly excites
  13. L18:2311 weakly excites
  14. L21:1010 weakly excites
  15. L8:1288 weakly suppresses
  16. L20:1445 weakly excites
  17. L12:559 weakly excites
  18. L17:3049 weakly excites
  19. L13:150 weakly suppresses
  20. L16:2035 weakly excites 'of' after a noun phrase
  21. L15:801 weakly excites 'of' inside organisation names
  22. L15:2521 weakly suppresses
  23. L15:969 weakly suppresses
  24. L15:817 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 -1.30 and -1.161 values between -1.16 and -1.021 values between -1.02 and -0.882 values between -0.88 and -0.745 values between -0.74 and -0.6115 values between -0.61 and -0.4732 values between -0.47 and -0.33188 values between -0.33 and -0.19847 values between -0.19 and -0.061869 values between -0.06 and 0.08753 values between 0.08 and 0.22137 values between 0.22 and 0.3622 values between 0.36 and 0.506 values between 0.50 and 0.631 values between 0.63 and 0.774 values between 0.77 and 0.912 values between 0.91 and 1.055 values between 1.05 and 1.191 values between 1.19 and 1.322 values between 1.32 and 1.462 values between 1.60 and 1.742 values between 1.74 and 1.885 values between 1.88 and 2.012 values between 2.15 and 2.291 values between 2.70 and 2.841 values between 3.12 and 3.251 values between 4.08 and 4.22offnegativepositiveoften

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. Jvm
  2. heimer
  3. ej
  4. "]}⏎
  5. 经
  6. flexDirection
  7. aporation
  8. 炙
  9. avar
  10. ens

Suppresses

  1. toile
  2. 五星
  3. eral
  4. clases
  5. ths
  6. -stock
  7. thest
  8. шло
  9. [class
  10. 遑