Neuron Explorer

Qwen3-0.6B

second word of 'Centers for Disease'

A neuron in layer 4 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 middle word of a name that is already half written, most often 'Disease' in 'Centers for Disease Control'. It seems to help the model finish a known name. When it was turned up, the written text did not show any clear theme.

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

    World War (1917-1918), the Second World War (1941-45), and events such as the Cuban Missile Crisis (1963).⏎During the Seminole Wars, the Spanish

  2. #2 peak

    number of connected in a tiny lobule. A minor salivary gland<|endoftext|>Applying Epidemiology To Program Design For Chronic Disease⏎According to the Centers for Disease Control and Prevention, 6 of the top 7 causes of death in the

  3. #3 peak

    osis. Supplemental vitamin D is popular because it is difficult (if not impossible) to get enough from food. In addition, as noted by the Centers for Disease Control and Prevention (CDC), our bodies

  4. #4 peak

    are more vulnerable to the Hepatitis C virus.⏎No vaccine is available for Hepatitis C, however it is curable with short term treatment.⏎The Centers for Disease Control and Prevention partners with community health organizations to help contain the infection. The AP

  5. #5 peak

    % quit rate." These statements are contrary to findings from two recent independent studies. But try telling that to U.S. health officials.⏎According to the Centers for Disease Control (CDC), the ALA's N-O-T program is the only smoking

  6. #6 peak

    as cases of COVID-19 continue to spike, you may be wondering when to get vaccinated against both viruses.⏎The answer is: Now.⏎The Centers for Disease Control (CDC) urges you--and anyone from six months of age or older

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 have been thinking about lately, and I have been thinking about lately. So, this is the part of the whole. Okay, so, I am a part of the whole. So, this is the part of the whole. Okay

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

    who cared?)⏎So the sample results can be matched to the individual. Wired up to the internet the results can be fed in real time to the appropriate medical practitioner

  2. #2 peak

    1890. After appearance of troops at Pine Ridge, he advocated for peace with the European Americans⏎Date de dernière mise à jour

  3. #3 peak

    .org.uk/10.1177%2F1359104505053754<|endoftext|>From Sioux Falls to Rapid City, fire weather meteorologists are watching conditions closely. And until we

  4. #4 peak

    range. By early spring, when most of the range and the crests of the adjacent San Gabriel range have lost their kiss of snow, big San Gorgonio often still shines white—the one true jewel

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. L3:50 strongly suppresses
  2. L3:691 suppresses
  3. L3:456 excites
  4. L2:41 excites
  5. L3:59 suppresses
  6. L3:464 suppresses
  7. L3:272 suppresses
  8. L3:421 excites
  9. L2:1710 excites
  10. L3:2140 excites
  11. L2:111 excites
  12. L3:115 excites
  13. L3:187 suppresses
  14. L0:1224 excites
  15. L0:178 excites
  16. L3:122 suppresses
  17. L0:1301 excites
  18. L2:2328 excites
  19. L2:2692 suppresses
  20. L2:1657 suppresses
  21. L2:1899 suppresses
  22. L2:70 suppresses
  23. L2:1036 suppresses
  24. L2:356 suppresses

Feeds later layers

  1. L8:1 weakly suppresses
  2. L27:2326 weakly excites
  3. L27:1995 weakly excites
  4. L27:1665 weakly excites
  5. L27:1974 weakly excites
  6. L27:2211 weakly excites
  7. L27:2116 weakly excites unclear: first and boundary tokens
  8. L27:107 weakly excites
  9. L10:0 weakly suppresses
  10. L27:1371 weakly excites
  11. L5:36 weakly excites
  12. L6:116 weakly suppresses
  13. L5:72 weakly excites
  14. L25:5 weakly suppresses
  15. L27:194 weakly excites
  16. L27:549 weakly excites
  17. L27:259 weakly excites
  18. L16:125 weakly suppresses unclear: mid-word and mid-phrase fragments
  19. L26:1564 weakly excites
  20. L11:0 weakly suppresses
  21. L16:166 weakly suppresses
  22. L27:1490 weakly excites
  23. L16:186 weakly suppresses
  24. L26:120 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.

2 values between -3.39 and -3.251 values between -3.12 and -2.981 values between -2.84 and -2.701 values between -2.70 and -2.573 values between -2.57 and -2.433 values between -2.43 and -2.2910 values between -2.29 and -2.155 values between -2.15 and -2.029 values between -2.02 and -1.8816 values between -1.88 and -1.7417 values between -1.74 and -1.6024 values between -1.60 and -1.4757 values between -1.47 and -1.3383 values between -1.33 and -1.19108 values between -1.19 and -1.05152 values between -1.05 and -0.92218 values between -0.92 and -0.78264 values between -0.78 and -0.64352 values between -0.64 and -0.50360 values between -0.50 and -0.37363 values between -0.37 and -0.23426 values between -0.23 and -0.09754 values between -0.09 and 0.05391 values between 0.05 and 0.18166 values between 0.18 and 0.3254 values between 0.32 and 0.4617 values between 0.46 and 0.6011 values between 0.60 and 0.736 values between 0.73 and 0.8712 values between 0.87 and 1.015 values between 1.01 and 1.153 values between 1.15 and 1.283 values between 1.28 and 1.422 values between 1.42 and 1.562 values between 1.56 and 1.702 values between 1.70 and 1.832 values between 1.83 and 1.973 values between 1.97 and 2.11offnegativepositiveoften

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. Packages
  2. Tet
  3. Emp
  4. Sheets
  5. Symbols
  6. Eternal
  7. Wrest
  8. ücken
  9. agrid
  10. Gew

Suppresses

  1. 自身的
  2. 本身就是
  3. 本身的
  4. 的基础
  5. yourself
  6. 本身
  7. 的话
  8. 的一
  9. license
  10. 一趟