Neuron Explorer

Qwen3-0.6B

'of' after effectiveness or instead

A neuron in layer 2 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 'of' in phrases like 'the effectiveness of', 'instead of' and 'the nutritional value of', where a noun is about to be followed by what it applies to. It seems to be less active on sentence-linking words like 'Also' and 'Additionally'. Turning it up or down changed the writing style but not the topic in any clear way.

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

    . Instead of removing the tumor, stereotactic radiotherapy damages the DNA of tumor cells.

  2. #2 peak

    M⏎Am Ind Hyg Assoc J 1980 Oct; 41(10):758-765⏎The effectiveness of open versus closed-face cassettes in aerosol sampling was compared. Thirty breathing

  3. #3 peak

    the number of inefficient vehicles on the road, thereby reducing GHG emissions, and can also prompt car manufacturers to provide more efficient vehicles to the public. The effectiveness of these programs has been demonstrated in high

  4. #4 peak

    a shift in thinking about renewable energy storage. Instead of generating electricity first and then converting the energy for storage, the idea is to use

  5. #5 peak

    compacted or poorly drained. Instead of treating the symptoms using pesticides, it’s important to discover and treat the reason the

  6. #6 peak

    ugreek plant are also edible and are often used in cooking, particularly in Indian and Middle Eastern cuisines. In this article, we will explore the nutritional value of fenugreek leaves per 100g and determine whether they are healthy

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: ofs the of the and ands, and and,, e y y y y y y, y y y y e, y ( o (, o e-es, t o T t t teso ( e (

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

    can help you create custom content for your target audience. Additionally, AI can help you create more engaging blog posts by automatically generating ideas based on your data sources. Additionally, AI writing assistants can also help improve your writing quality by suggesting improvements to your

  2. #2 peak

    can be applied<|endoftext|>The Federal Scientist: Multiple Roles and Moral Issues⏎This scenario highlights potential dilemmas encountered by postdoctoral fellows in a research setting. This scenario looks at the multiple roles a scientist must play and how it can affect decision making

  3. #3 peak

    one of the richest sources for many nutrients. Also it is very high in minerals and vitamins, such as vitamin A, vitamin B, folate and iron. Also they contain a bit protein inside them and they contain a very healthy dose of antioxidants

  4. #4 peak

    , depict rock type, distribution, properties and its relative age. They provide information about the Earth’s structure and provide a baseline for data related to energy resources, mineral resources, natural hazards, water

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. L1:104 suppresses
  2. L0:342 excites
  3. L1:272 excites
  4. L0:1301 suppresses
  5. L1:480 weakly excites
  6. L0:178 weakly suppresses
  7. L0:1224 weakly suppresses
  8. L1:17 weakly suppresses
  9. L1:742 weakly excites
  10. L0:104 weakly excites
  11. L1:1621 weakly suppresses
  12. L1:494 weakly suppresses
  13. L1:647 weakly suppresses
  14. L1:176 weakly suppresses
  15. L0:2142 weakly suppresses
  16. L1:3061 weakly suppresses
  17. L1:251 weakly suppresses
  18. L0:2984 weakly suppresses
  19. L0:749 weakly excites
  20. L1:1348 weakly excites
  21. L0:1199 weakly excites
  22. L0:2540 weakly suppresses
  23. L0:2010 weakly suppresses
  24. L1:1487 weakly suppresses

Feeds later layers

  1. L8:168 excites
  2. L16:798 excites
  3. L6:1458 excites
  4. L17:764 excites
  5. L8:1462 suppresses
  6. L8:642 excites
  7. L4:2 suppresses
  8. L16:2610 excites
  9. L4:5 excites
  10. L7:478 weakly excites
  11. L10:222 weakly excites
  12. L17:1554 weakly excites
  13. L3:56 weakly suppresses
  14. L12:594 weakly suppresses
  15. L8:212 weakly suppresses
  16. L6:1538 weakly excites
  17. L15:510 weakly suppresses
  18. L15:2274 weakly suppresses
  19. L17:1701 weakly suppresses
  20. L7:883 weakly excites
  21. L8:227 weakly excites
  22. L4:323 weakly excites
  23. L15:2952 weakly suppresses
  24. L15:910 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 -2.44 and -2.362 values between -1.91 and -1.841 values between -1.46 and -1.392 values between -1.39 and -1.311 values between -1.31 and -1.245 values between -1.08 and -1.012 values between -1.01 and -0.935 values between -0.93 and -0.866 values between -0.86 and -0.788 values between -0.78 and -0.717 values between -0.71 and -0.638 values between -0.63 and -0.5613 values between -0.56 and -0.4822 values between -0.48 and -0.4134 values between -0.41 and -0.3352 values between -0.33 and -0.2683 values between -0.26 and -0.18238 values between -0.18 and -0.11673 values between -0.11 and -0.031897 values between -0.03 and 0.05592 values between 0.05 and 0.12179 values between 0.12 and 0.2053 values between 0.20 and 0.2716 values between 0.27 and 0.355 values between 0.35 and 0.423 values between 0.50 and 0.57offnegativepositiveoften

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. @Table
  2. otten
  3. @Api
  4. 境内
  5. 起了
  6. @Resource
  7. etable
  8. alyzed
  9. fertile
  10. cedes

Suppresses

  1. ழ
  2. ka
  3. 款
  4. past
  5. 型
  6. mode
  7. ko
  8. pass
  9. _mode
  10. cont