Neuron Explorer

Qwen3-0.6B

'this' pointing at the current document

A neuron in layer 20 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 'this' when it refers back to the page or document being read, as in 'read the rest of this entry' or 'at the end of this post'. Turned up, the model wrote text about cars, weather and houses with no clear link, so steering gives no support.

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 IDEAS help page. Note that these files are not on the IDEAS site. Please be patient as the files may be large.⏎As the access to this document is restricted, you may want to look for

  2. #2 peak

    the chance to explore STEM topics in a way that deepens their knowledge while also developing 21st-century skills.⏎Read the rest of this entry »⏎Filed under: Gaming, Services, Software, 3 million dollar

  3. #3 peak

    of its “Fusion Engine” by 2019 after which licensees will begin building commercial models by 2022.⏎If you liked this article, please give it a quick review on ycombinator or Stumble

  4. #4 peak

    not worry about the Digest/ Guide because after using our Pdf solutions you will never find need of them. You will get all the Pdf files at the end of this post, so we would like to request you to read full article.⏎|Mah

  5. #5 peak

    Islands is available on the Solomon Islands Country Page and from other Department of State publications and other sources listed at the end of this fact sheet.⏎U.S.-SOLOMON ISLANDS RELATIONS⏎During

  6. #6 peak

    the design, installation, maintenance, and inspection of all chimneys, fireplaces, venting systems, and solid-fuel-burning appliances. The reach of this standard is vast, as it can be

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: the topic of the subject matter of the project, and the subject matter of the project material. In the project material material, material, and so on. I would be using materials in the project material to project the topic material content material,

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

    70s. Providing comparative analysis as well as reflection on the evolution of this important genre, Urban Chroniclers in Modern Latin America is the first systematic study of the Latin American writers who forged a new reading public in the early twentieth century.<|endoftext|>

  2. #2 peak

    research is suggesting that children who eat more foods with gluten before the age of two have a greater risk of developing celiac disease if they are genetically predisposed to the condition.⏎This study provides some insight into why some children who are genetically predisposed

  3. #3 peak

    play a major role in the etiology of alcoholism in women. Women should be well represented in the efforts currently under way to elucidate the molecular basis of the genetic susceptibility to alcoholism.(JAMA. 1992;2

  4. #4 peak

    of Dementia or Alzheimer’s Disease⏎Alzheimer’s Disease is a leading cause of death in the United States, with 5.8M people suffering from the disease annually. The disease falls under the category of symptoms referred to as dementia,

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. L19:1122 strongly suppresses
  2. L19:705 strongly suppresses
  3. L16:715 strongly excites
  4. L19:1496 strongly suppresses
  5. L19:673 strongly excites
  6. L19:2198 suppresses
  7. L14:2291 suppresses
  8. L19:2711 suppresses
  9. L11:810 suppresses
  10. L18:2142 excites
  11. L17:390 excites
  12. L19:604 suppresses
  13. L15:2449 excites line breaks mid-sentence
  14. L19:207 suppresses
  15. L14:990 suppresses
  16. L19:2183 suppresses
  17. L13:2099 suppresses
  18. L18:305 suppresses
  19. L17:1027 excites
  20. L18:1764 excites
  21. L18:848 suppresses
  22. L13:988 excites
  23. L17:1467 excites
  24. L19:2490 excites

Feeds later layers

  1. L21:831 strongly suppresses
  2. L23:355 strongly suppresses
  3. L21:843 strongly suppresses
  4. L21:68 excites
  5. L21:323 suppresses
  6. L25:525 suppresses
  7. L27:735 suppresses
  8. L22:811 suppresses
  9. L25:390 suppresses
  10. L22:207 excites the prefix 'non' before a word
  11. L22:359 suppresses
  12. L21:498 suppresses
  13. L27:227 suppresses
  14. L21:368 suppresses
  15. L22:2089 suppresses
  16. L21:2748 suppresses
  17. L26:510 suppresses
  18. L25:444 weakly excites
  19. L22:387 weakly suppresses
  20. L21:1402 weakly excites
  21. L22:449 weakly excites
  22. L26:359 weakly suppresses
  23. L21:432 weakly suppresses
  24. L27:521 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 -8.84 and -8.381 values between -7.93 and -7.471 values between -6.56 and -6.113 values between -6.11 and -5.652 values between -5.65 and -5.205 values between -5.20 and -4.741 values between -4.74 and -4.295 values between -4.29 and -3.8310 values between -3.83 and -3.3813 values between -3.38 and -2.9313 values between -2.93 and -2.4732 values between -2.47 and -2.0241 values between -2.02 and -1.5674 values between -1.56 and -1.11109 values between -1.11 and -0.65222 values between -0.65 and -0.201447 values between -0.20 and 0.261441 values between 0.26 and 0.71376 values between 0.71 and 1.1738 values between 1.17 and 1.6212 values between 1.62 and 2.0816 values between 2.08 and 2.538 values between 2.53 and 2.996 values between 2.99 and 3.444 values between 3.44 and 3.909 values between 3.90 and 4.351 values between 4.35 and 4.812 values between 4.81 and 5.261 values between 5.26 and 5.715 values between 5.71 and 6.172 values between 6.17 and 6.624 values between 6.62 and 7.081 values between 7.99 and 8.441 values between 8.90 and 9.35offnegativepositiveoften

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. <title
  2. wich
  3. .dds
  4. unga
  5. .Views
  6. loid
  7. resse
  8. .cgi
  9. Franç
  10. resta

Suppresses

  1. 事项
  2. 項目
  3. 项目的
  4. 项目
  5. 因素
  6. situation
  7. phenomenon
  8. statements
  9. 事件
  10. 场合