Neuron Explorer

Qwen3-0.6B

'to' after 'as opposed' or 'related'

A neuron in layer 0 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 reacts to the word 'to' when it follows phrases like 'as opposed', 'related', or 'akin', as in 'as opposed to'. Turning it up did not produce any obvious change in topic or style that is easy to describe.

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

    cachexia. However, these compounds are Food and Drug Administration approved, have known safety and efficacy, and predictable potency and purity profiles. These formulations are unrelated to synthetic cannabinoids and the products that contain them.⏎Today, companies buy these synthetic cannabinoids

  2. #2 peak

    ival home in 1870 at 424 Fifth Street Southeast. Related to Woodbury Fisk by marriage, Thomas Andrews built his Italianate Revival home

  3. #3 peak

    pounds in a healthier manner, it is perfect to burn fat gradually. And incredibly, most professionals consent which you can accomplish by using a diet regime. As opposed to a extreme life-style change, the secret to success is helping to make

  4. #4 peak

    to sovereigns or states; a confederate.⏎- (v.) Anything associated with another as a helper; an auxiliary.⏎- (v.) Anything akin to another by structure, etc.⏎- (v. t.) To unite, or

  5. #5 peak

    them,” said Fant. “But when this is not possible, these plant collections are maintained by continually crossing with other plants to produce new seed. This is akin to animals in zoo collections. Zoos have used genetic information to develop ‘studbooks

  6. #6 peak

    learning has a solid feeling of hearing and is delicate to sound. This successful learning strategy effectively retains data and comprehends the approaching boost through the ear. As opposed to the visual sort, hear-able sort individuals will think that its hard to gather

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 don't have any idea to do, I think the answer is to do. I think the answer is to do. So, I have I think and I need I'm I don I'm I don I'm I don... I

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

    according to the World Health Organization (WHO). In the United States 110

  2. #2 peak

    according to an article in the January 2006 issue of AGD Impact

  3. #3 peak

    according to Takei.<|endoftext|>Do you want to see the dazzling display of the Per

  4. #4 peak

    according to the U.S. Drug Enforcement Administration (DEA).⏎Ketamine Found

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

None: this is the first layer.

Feeds later layers

  1. L6:338 suppresses
  2. L1:640 suppresses
  3. L13:3 suppresses
  4. L8:575 weakly suppresses
  5. L20:2995 weakly suppresses
  6. L21:2793 weakly suppresses
  7. L7:15 weakly suppresses
  8. L27:2553 weakly suppresses capital T starting a word
  9. L7:7 weakly suppresses
  10. L8:506 weakly suppresses
  11. L3:1420 weakly suppresses
  12. L7:8 weakly suppresses
  13. L19:442 weakly suppresses
  14. L5:709 weakly suppresses
  15. L3:2883 weakly suppresses
  16. L7:23 weakly excites
  17. L1:2326 weakly suppresses
  18. L8:1918 weakly suppresses
  19. L19:1363 weakly suppresses
  20. L8:705 weakly suppresses
  21. L21:703 weakly suppresses
  22. L2:2601 weakly suppresses
  23. L2:1984 weakly suppresses
  24. L15:2012 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 -2.65 and -2.531 values between -1.94 and -1.821 values between -1.46 and -1.341 values between -1.34 and -1.232 values between -1.23 and -1.111 values between -0.99 and -0.871 values between -0.87 and -0.755 values between -0.75 and -0.635 values between -0.63 and -0.516 values between -0.51 and -0.4015 values between -0.40 and -0.2871 values between -0.28 and -0.16395 values between -0.16 and -0.043029 values between -0.04 and 0.08305 values between 0.08 and 0.2025 values between 0.20 and 0.329 values between 0.32 and 0.439 values between 0.43 and 0.556 values between 0.55 and 0.675 values between 0.67 and 0.795 values between 0.79 and 0.911 values between 0.91 and 1.031 values between 1.03 and 1.141 values between 1.14 and 1.263 values between 1.26 and 1.381 values between 1.38 and 1.501 values between 1.74 and 1.861 values between 1.86 and 1.971 values between 1.97 and 2.09offnegativepositiveoften

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. �
  2. ationally
  3. icking
  4. icks
  5. .Attribute
  6. ick
  7. rew
  8. Clem
  9. 下行
  10. 更多

Suppresses

  1. 给了我
  2. 范文
  3. stock
  4. оказыва
  5. 给他们
  6. 雅黑
  7. Sunni
  8. "','
  9. stocks
  10. ẫu