Neuron Explorer

Qwen3-0.6B

"of" after an agency or department name

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 lights up on the word "of" when it follows the name of a government department or institution, such as "Department of Natural Resources". Turning it up made the model write about unrelated subjects, so the steering does not show 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

    of Chippewa Indians won fishing rights in 1990 ceded by a 1837 treaty and works with the Minnesota Department of Natural Resources to set limits that preserves the fishery. Indian fishing rights issues in

  2. #2 peak

    kin Uses Technology, Science to Treat Troubled Youth⏎From the October 2013 Desktop News | Troubled youth in the custody of the Alabama Department of Youth Services have been given a second chance at success thanks to the work of Dr

  3. #3 peak

    of success,” Rees says.⏎State officials also connect the recent decline in adult sturgeon numbers to sea lions. But Tony Nigro with Oregon’s Department of Fish and

  4. #4 peak

    of arsenic, cadmium, manganese, and lead and birth defects prevalence. The studies used a statewide database of private well contaminants collected by the North Carolina Department of Health and Human Services Division of Public Health as well as data from the North Carolina

  5. #5 peak

    res rouse him? The<|endoftext|>Juvenile striped bass spawned in the Chesapeake Bay were at a record low at the end of the summer. The Maryland Department of Natural Resources recently completed its annual young of the year survey and found far fewer juvenile

  6. #6 peak

    sensible cooling capacity and latent cooling capacity.⏎Sensible cooling capacity is the air conditioner's ability to remove heat and lower temperature<|endoftext|>Last month, the National Academies of Science, Engineering, and Medicine hosted a 3-day workshop: "Optim

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: "I am a student of the art of the student, and I am in the middle of this very thing." What is being said here, and what I am doing. Let's see. If I were to explain this to someone,

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

    , including issues with heart health, kidney problems, and seizures.⏎A person suffering from anorexia nervosa will typically fall under one of two forms of self-starvation: restricting, in which the person participates in extreme diets, fasting, or

  2. #2 peak

    make sure you comprehend the question, which is needed to be enlightened in your paper. You need to evaluate the topic, which will take some time. World War 2 is such a broad theme that in order to write a profound essay, you

  3. #3 peak

    eltis polypodioides. It gets its name from its effective, defense mechanism against droughts. Being able to survive decades without water and still be able to return to a healthy state after some watering. This has made a lot of

  4. #4 peak

    ago, the Navajo Nation and Southern Ute tribes languished side by side, mired in high unemployment and poverty.⏎Today, worth billions, the Southern Ute Indian Tribe is one of the richest in the United

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. L8:119 suppresses
  2. L8:2854 excites
  3. L7:266 suppresses
  4. L8:204 suppresses
  5. L8:177 suppresses
  6. L8:149 suppresses
  7. L8:182 excites
  8. L6:91 excites
  9. L8:376 suppresses
  10. L7:26 suppresses
  11. L3:4 excites
  12. L7:23 suppresses
  13. L8:941 suppresses
  14. L8:320 excites
  15. L8:282 excites
  16. L7:6 weakly suppresses
  17. L8:230 weakly suppresses
  18. L4:559 weakly suppresses
  19. L3:0 weakly excites
  20. L8:25 weakly suppresses
  21. L8:296 weakly excites 'of' inside fixed noun phrases
  22. L8:71 weakly excites
  23. L7:57 weakly excites
  24. L4:274 weakly excites

Feeds later layers

  1. L10:308 strongly suppresses
  2. L10:146 excites
  3. L11:341 suppresses
  4. L11:222 excites
  5. L11:25 excites
  6. L10:642 suppresses
  7. L11:285 suppresses
  8. L10:137 excites 'the' before a noun after 'of'
  9. L11:1110 suppresses
  10. L10:208 suppresses
  11. L10:202 excites
  12. L11:180 suppresses
  13. L14:617 suppresses
  14. L12:219 weakly suppresses
  15. L11:2138 weakly suppresses
  16. L10:703 weakly suppresses
  17. L10:498 weakly excites hyphen inside numbers
  18. L10:1132 weakly suppresses
  19. L13:2661 weakly suppresses
  20. L10:620 weakly suppresses middle of technical terms in science text
  21. L12:2 weakly suppresses
  22. L16:176 weakly suppresses
  23. L14:1780 weakly excites
  24. L21:2941 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 -2.99 and -2.821 values between -2.82 and -2.661 values between -2.66 and -2.491 values between -2.33 and -2.161 values between -2.16 and -1.991 values between -1.99 and -1.835 values between -1.83 and -1.664 values between -1.66 and -1.508 values between -1.50 and -1.3313 values between -1.33 and -1.1613 values between -1.16 and -1.0020 values between -1.00 and -0.8334 values between -0.83 and -0.6658 values between -0.66 and -0.5099 values between -0.50 and -0.33253 values between -0.33 and -0.171245 values between -0.17 and 0.001447 values between 0.00 and 0.17389 values between 0.17 and 0.33131 values between 0.33 and 0.5067 values between 0.50 and 0.6743 values between 0.67 and 0.8318 values between 0.83 and 1.0019 values between 1.00 and 1.1611 values between 1.16 and 1.331 values between 1.33 and 1.506 values between 1.50 and 1.667 values between 1.66 and 1.833 values between 1.83 and 2.001 values between 2.00 and 2.163 values between 2.16 and 2.331 values between 2.33 and 2.492 values between 3.16 and 3.321 values between 3.49 and 3.66offnegativepositiveoften

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. eties
  2. ité
  3. encion
  4. iences
  5. ientes
  6. erts
  7. entially
  8. erà
  9. atches
  10. AttributeSet

Suppresses

  1. quam
  2. шение
  3. ibus
  4. ucus
  5. uke
  6. tica
  7. 边界
  8. fühl
  9. ux
  10. xDB