Neuron Explorer

Qwen3-0.6B

unclear: periods inside numbers and names

A neuron in layer 9 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 one reacts to periods that sit inside things like version numbers, IP-style numbers, and some names or abbreviations, and it is most active on the dot itself. It is mixed, though, and the steering outputs show no clear theme. Turning it up made the text repetitive and off topic.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    a list of a single compound type like Entity or Structure, was already supported before 9.0.⏎Until 188.8.131.52, if you wanted to create a list of users, you would create

  2. #2 peak

    NSM2012. You'll have 30 days to complete the course.⏎For the younger driver

  3. #3 peak

    be (at time of writing) 22.214.171.124-kgpd-071097. This is the version also known as 0.5.0. If you

  4. #4 peak

    with symptoms such as numbness and tingling in the legs and feet, may cause loss of strength in the muscles it stimulates, depending on the degree of nerve compression.⏎Sciatica occurs when the bones in the spine press on the sciatic nerve

  5. #5 peak

    taking their pills as directed, the study found.⏎“It does appear that there is a modest reduction in cancer among middle-aged and older men,” said Dr. J. Michael Gaziano, chief of the division of aging at Brigham and Women’s Hospital

  6. #6 peak

    and other features.⏎Merriam-Webster Medical Dictionary⏎Sponsored by the National Library of Medicine, this site includes⏎links to encyclopedias, glossaries,

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 the The following, I have been thinking about the The following, I have I have been thinking about lately, I have been thinking about I have been thinking about I have been thinking about I have been thinking about 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

    or and donor layers, or domains, in polymer-based solar cells could increase their efficiency, according to a team of researchers that included physicists from North Carolina State University. Their findings shed light on the inner workings of these solar cells, and could lead

  2. #2 peak

    Maths Chapter 6 Exercise 6.5, we have discussed the Pythagoras theorem. Other important topics related to the chapter are covered in the following exercises.<|endoftext|>What you will learn in this module⏎Understand different tasks that have to

  3. #3 peak

    Solid Shapes⏎✔ Telling Time⏎✔ Comparing Numbers⏎Why They’ll Love It⏎Your students will review this year’s math concepts in a fun way! They will get:⏎☆ summertime jokes⏎☆ knock knock jokes⏎☆ cell

  4. #4 peak

    related descendants of that dog to fix the traits they want, said Dr. Donald F. Patterson, a professor of medicine and medical genetics at the University of Pennsylvania. If a popular sire also carries a genetic defect, it too spreads rapidly through the

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. L5:155 suppresses
  2. L4:7 excites
  3. L4:1546 suppresses
  4. L6:2302 suppresses 'and' joining two parallel items
  5. L4:848 excites
  6. L5:168 suppresses
  7. L6:305 suppresses
  8. L8:309 suppresses
  9. L6:1736 suppresses
  10. L2:9 suppresses sentence-final periods vs dots in code
  11. L6:466 excites
  12. L8:143 weakly suppresses
  13. L8:304 weakly suppresses
  14. L1:130 weakly suppresses
  15. L8:740 weakly excites
  16. L6:871 weakly suppresses
  17. L8:381 weakly excites
  18. L8:174 weakly suppresses
  19. L8:2014 weakly excites
  20. L6:327 weakly excites
  21. L5:2159 weakly suppresses
  22. L8:1848 weakly excites
  23. L6:2399 weakly suppresses
  24. L0:1475 weakly excites

Feeds later layers

  1. L11:1025 suppresses
  2. L17:1105 suppresses
  3. L16:894 suppresses full stops ending informational sentences
  4. L11:132 suppresses period ending an opening topic sentence
  5. L14:153 suppresses
  6. L14:155 suppresses sentence-ending periods in prose
  7. L21:1086 suppresses end of sentence before a claim or statistic
  8. L16:1545 suppresses
  9. L19:344 suppresses periods ending explanations
  10. L12:1062 suppresses
  11. L22:67 suppresses
  12. L18:791 suppresses
  13. L16:436 suppresses
  14. L17:200 suppresses
  15. L24:353 suppresses decimal point inside a number
  16. L19:433 suppresses
  17. L17:347 suppresses full stops ending informative sentences
  18. L17:1772 suppresses
  19. L13:9 excites
  20. L18:1437 suppresses
  21. L10:717 suppresses
  22. L13:345 suppresses
  23. L16:797 excites
  24. L20:505 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 -5.22 and -5.071 values between -5.07 and -4.911 values between -4.60 and -4.452 values between -3.98 and -3.822 values between -3.82 and -3.671 values between -3.67 and -3.513 values between -3.51 and -3.363 values between -3.36 and -3.203 values between -3.20 and -3.052 values between -3.05 and -2.892 values between -2.89 and -2.732 values between -2.73 and -2.585 values between -2.58 and -2.429 values between -2.42 and -2.278 values between -2.27 and -2.113 values between -2.11 and -1.9610 values between -1.96 and -1.809 values between -1.80 and -1.655 values between -1.65 and -1.498 values between -1.49 and -1.3314 values between -1.33 and -1.189 values between -1.18 and -1.027 values between -1.02 and -0.878 values between -0.87 and -0.718 values between -0.71 and -0.5617 values between -0.56 and -0.4020 values between -0.40 and -0.2582 values between -0.25 and -0.09446 values between -0.09 and 0.07881 values between 0.07 and 0.221176 values between 0.22 and 0.38783 values between 0.38 and 0.53297 values between 0.53 and 0.6970 values between 0.69 and 0.8410 values between 0.84 and 1.00offnegativepositiveoften

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. og
  2. id
  3. oph
  4. 引用日期
  5. encore
  6. ro
  7. .pipeline
  8. idia
  9. puedo
  10. imiters

Suppresses

  1. 然而
  2. 而在
  3. however
  4. 这其中
  5. whereas
  6. holders
  7. tery
  8. 而对于
  9. however
  10. 但这