Neuron Explorer

Qwen3-0.6B

'to' after 'prior' (before)

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 word 'to' in the phrase 'prior to', as in 'prior to 1970'. It was almost silent on most other text. When it was turned up, the written text did not clearly change in any consistent 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

    opportunity for you to learn more about new cultures, and explore some of the diversity that makes the world such a wonderful place.⏎Global Diversity Awareness Month timeline⏎Prior to and throughout the 2014 Olympic Games in So

  2. #2 peak

    are pondering how to make your computer virus, you’ve come towards the right place. There are many tutorials and steps on making a malware. However , prior to this, it is important to understand what the pathogen is, just how it

  3. #3 peak

    helps cut down on friction, making it useful for keeping skin dry and helping to prevent rashes.”⏎What’s the problem with baby powder with talc?⏎Prior to 1970, many powders contained talc which also contained asbestos

  4. #4 peak

    the researchers noted that prior to the existence of carpool lanes commuters said carpooling was too expensive. After

  5. #5 peak

    on CO<|endoftext|>Multilateral Trade Negotiations: Tips and Techniques (2013)⏎Duration of event:⏎Event focal point email:⏎Other events details:⏎Prior to the Uruguay Round of mult

  6. #6 peak

    96, the National⏎Science Education Standards (NSES) have been⏎at the center of the science education reform⏎movement in the United States. Prior to that⏎time, the National Science Foundation, other⏎government agencies, and private

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: What is the value of `x`? So, $ x$? Let me write an equation for $ x$. Let me write $ x$ in terms of $ x$. So, $ x$? So, $ x

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

    imates: Sow in late fall through winter for cool–season harvest.⏎When to Start Inside: Not recommended; transplant stress can cause plants to bolt (prematurely flower).⏎Days to Emerge: 5–10 days⏎

  2. #2 peak

    recognize that as long as it is accomplished and done on time, that it is okay. In the beginning, be prepared to patiently ask and answer a lot of questions! Why do emergency numbers need to be

  3. #3 peak

    cameras in hand and away from the fenced perimeter, be prepared to possibly be stopped and answer their questions regarding your visit. The staff housing located not

  4. #4 peak

    ’re working on⏎- Create a pre-comp to easily duplicate certain animations for re-use⏎Here’s another great example video done by Nick in After Effects using precomps!⏎Check back next week to learn the difference between frame rate and shutter speed

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. L3:2829 excites
  2. L3:691 weakly excites
  3. L3:187 weakly excites
  4. L3:59 weakly excites
  5. L3:50 weakly excites
  6. L2:2648 weakly suppresses
  7. L3:272 weakly excites
  8. L3:3030 weakly excites
  9. L3:512 weakly suppresses
  10. L2:1494 weakly suppresses
  11. L3:456 weakly suppresses
  12. L1:709 weakly suppresses
  13. L3:122 weakly excites
  14. L2:41 weakly suppresses
  15. L2:1292 weakly suppresses
  16. L1:2261 weakly suppresses
  17. L2:1277 weakly suppresses
  18. L1:2076 weakly excites
  19. L3:464 weakly excites
  20. L1:2922 weakly excites
  21. L0:1212 weakly suppresses
  22. L2:493 weakly suppresses
  23. L2:729 weakly suppresses
  24. L2:1690 weakly suppresses

Feeds later layers

  1. L9:2535 excites
  2. L8:1999 excites
  3. L13:2439 excites
  4. L20:2371 excites
  5. L15:2342 excites
  6. L7:2428 excites
  7. L19:1298 excites
  8. L26:1614 weakly suppresses
  9. L7:1692 weakly excites
  10. L18:562 weakly excites
  11. L18:2984 weakly excites
  12. L8:1477 weakly excites
  13. L12:2991 weakly excites
  14. L15:2163 weakly excites
  15. L13:1327 weakly excites
  16. L9:2910 weakly suppresses
  17. L9:619 weakly excites
  18. L9:3002 weakly suppresses
  19. L6:794 weakly excites
  20. L20:1546 weakly excites
  21. L11:2584 weakly excites
  22. L8:1916 weakly excites
  23. L18:1933 weakly excites
  24. L9:563 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.

5 values between -0.96 and -0.6377 values between -0.63 and -0.302672 values between -0.30 and 0.031134 values between 0.03 and 0.3711 values between 0.37 and 0.703 values between 0.70 and 1.034 values between 1.03 and 1.361 values between 3.69 and 4.021 values between 11.99 and 12.32offnegativepositiveoften

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. componentWill
  2. 准备好
  3. -facing
  4. -before
  5. bevor
  6. iores
  7. (grammarAccess
  8. before
  9. *</
  10. theless

Suppresses

  1. 密度
  2. 不理
  3. cern
  4. Vaults
  5. iceberg
  6. density
  7. 陸
  8. IMPLIED
  9. IMPLIED
  10. 各县