Neuron Explorer

Qwen3-0.6B

'well' inside a compound phrase

A neuron in layer 24 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 'well' when it joins another word into a phrase, like 'well-funded', 'well known' or 'well-being'. Turning it up didn't produce anything clearly tied to that idea, so the steering text is hard to read as a theme.

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

    with minefields and traps. Following Prime Minister Narendra Modi’s decision to strike at terrorists across the Line of Control in 2016, Pakistan responded with well-tested covert tools—ratcheting up the levels of violence against Indian forces in

  2. #2 peak

    ” of the justice system, including new technologies used in the courtroom, as well as court facilities and security, and watch a demonstration by Pocket, the dog whose well-trained nose aids Capitol Police in protecting the courthouse.⏎Thereafter, they get an

  3. #3 peak

    Clearinghouse for Comprehensive School Reform presents a suggested progression for the reform process. Whether you are just beginning to think about change or your school's reform efforts are well underway, Step-by-Step provides tools to assist you during this process.⏎State

  4. #4 peak

    this year (Nature, 2009). In 2015, Stephen Hawking and Russian billionaire Yuri Milner announced the start of a well-funded program, ‘Breakthrough Initiatives’, which aims to find extra-ter

  5. #5 peak

    , Fake News is a deliberate spread of misinformation, with the intent to mislead in order to gain financially or politically.⏎Fake News vs Satire⏎If a well known comedy programme featured a parody of a news presenter saying that Donald Trump’s hair

  6. #6 peak

    <|endoftext|>When it comes to workplace safety, few topics get as much attention as that of fire safety. Between fire drills and how to contain fires, most people are well aware of the fact that this is a significant risk. For many facilities, however

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: when all right to right all right to right all right to right all right all right right right right right right right right right right right right right right right right right right right right right right right right right right right right right right right right

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

    19th century, has added many decorative elements to the castle although retaining much of the military appearance of what was then the summer residence of the Marquis de Armijo.<|endoftext|>Thread: FNG methods of attack question⏎05-0

  2. #2 peak

    & Co”, and Alfred Deakin (St.⏎Kilda Cemetery), thrice Prime Minister of Australia. Residing at 7⏎Barnalo Grove, Armad<|endoftext|>🇮🇳 Further to

  3. #3 peak

    , so these children often grow up to be productive workers and rule-followers. The downside of this strategy is that they could grow up to be sheep-like and uncreative, requiring instructions from others to function, rather than feeling free to think for

  4. #4 peak

    195 square kilometres and enters Lake Huron at Point Clark. The Pine River sub-watershed is located in Bruce County, Township of Huron-Kinloss and includes the Village of Ripley.⏎The Bruce Peninsula Biosphere Association (

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. L19:134 weakly excites
  2. L18:178 weakly suppresses
  3. L17:1403 weakly suppresses
  4. L20:342 weakly suppresses
  5. L21:38 weakly excites
  6. L3:1066 weakly suppresses
  7. L17:2984 weakly excites blank before a number in prevalence claims
  8. L4:1587 weakly excites
  9. L9:2511 weakly excites
  10. L6:2675 weakly excites
  11. L5:3054 weakly excites 'to' after ease or safety adjectives
  12. L1:1133 weakly suppresses
  13. L23:619 weakly suppresses
  14. L11:1916 weakly suppresses
  15. L9:1408 weakly excites
  16. L12:2408 weakly suppresses
  17. L7:2412 weakly suppresses
  18. L19:2221 weakly excites
  19. L21:928 weakly suppresses
  20. L20:413 weakly suppresses
  21. L12:172 weakly excites
  22. L20:669 weakly suppresses
  23. L21:304 weakly excites
  24. L13:23 weakly excites

Feeds later layers

  1. L27:1484 weakly excites
  2. L27:572 weakly excites
  3. L26:880 weakly excites
  4. L27:2138 weakly suppresses
  5. L25:1749 weakly excites
  6. L25:670 weakly excites
  7. L25:1866 weakly excites
  8. L25:1714 weakly suppresses
  9. L25:2340 weakly suppresses 'to' after a verb, before 'avoid'
  10. L27:757 weakly excites
  11. L26:789 weakly excites
  12. L25:569 weakly excites
  13. L26:1879 weakly suppresses
  14. L26:2437 weakly suppresses
  15. L27:1296 weakly suppresses
  16. L25:1582 weakly suppresses
  17. L25:2353 weakly excites
  18. L25:1692 weakly excites
  19. L27:1836 weakly suppresses
  20. L27:672 weakly suppresses
  21. L25:1530 weakly suppresses
  22. L25:2593 weakly suppresses
  23. L25:753 weakly suppresses
  24. L25:1527 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.

2 values between -15.96 and -14.492 values between -14.49 and -13.011 values between -13.01 and -11.541 values between -11.54 and -10.075 values between -10.07 and -8.596 values between -8.59 and -7.126 values between -7.12 and -5.6420 values between -5.64 and -4.1725 values between -4.17 and -2.7085 values between -2.70 and -1.222871 values between -1.22 and 0.25773 values between 0.25 and 1.7361 values between 1.73 and 3.2022 values between 3.20 and 4.6811 values between 4.68 and 6.158 values between 6.15 and 7.622 values between 7.62 and 9.103 values between 9.10 and 10.572 values between 10.57 and 12.051 values between 29.73 and 31.211 values between 41.53 and 43.00offnegativepositiveoften

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. -offs
  2. ows
  3. -off
  4. Herz
  5. .off
  6. acon
  7. off
  8. ULATION
  9. bac
  10. ometown

Suppresses

  1. eti
  2. interest
  3. eno
  4. Interest
  5. loth
  6. *e
  7. setObject
  8. interest
  9. 箸
  10. Nor