Neuron Explorer

Qwen3-0.6B

closing brackets and parentheses

A neuron in layer 22 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the closing bracket, parenthesis, or full stop that ends a short aside or symbol, such as the ')' after '(Cádiz' or the '(%)' after a symbol. The pattern is common but loose, and turning it up or down did not give a clear theme.

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

    It stands for the number of expressions to follow. Each expression consists of a single line containing a positive integer number, an operator (one of +, - and *) and the second positive integer number. Every number has at most 500

  2. #2 peak

    places, in Sicily and Corsica, in Spain at Tartessus, and even beyond the pillars of Hercules at Gadeira (Cádiz)".⏎In the time of King

  3. #3 peak

    sidetrack⏎English → English⏎, v. i. [imp. & p. p. Departed⏎; p. pr. & vb.⏎.] [OE. departen to divide, part, depart, F.⏎d['

  4. #4 peak

    ued by||Banco Nacional de Cuba|⏎|Size||157 x 66 mm|⏎|Printer||American Bank Note Company (U.S.A.) – ABNC|⏎|Signatures||Ernesto (Che) Gue

  5. #5 peak

    makes sense for a function to return multiple things. In such cases, it's common to just return them in a tuple: "return (count, new_obj)".⏎Sometimes you might want to return objects of different types based on whether an operation

  6. #6 peak

    to rhyme with the word 'eye'. But, for whatever reason, it is usually pronounced as '/iy/' or “ee”.⏎How do you pronounce daemon in Golden Compass?⏎http://

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: The way people are supposed to think about the way people are supposed to think about themselves. All of the way across all of history, all of the way across all of the lengths of the length of the length of all of the lengths.

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 other meat) is submerged in a salty liquid (“brine”) and left to marinate to add flavour, tenderise AND add moisture so the chicken stays juicier when roasted.⏎Is raw chicken OK in the fridge for 5 days

  2. #2 peak

    is of even greater concern when considering the changing demographics.⏎The intention to perform a behavior is the most important factor to actually engaging in future behavior, according to Fishbein and Ajzens (Fishbein & Ajzen 1

  3. #3 peak

    troops. Officers inspect an howitzer. The troops march. German troops and officers eat. Soldiers listen to the radio. Troops put on uniforms. German JU-52 aircraft taxi and take off. A fleet of aircraft in flight.

  4. #4 peak

    ard apple, carissa, dialium, flacourtia (Indian plum), marula, vangueria, tamarind, vitex and ‘jujube’ are picked in the wild.⏎They are used as accompaniments

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. L0:110 excites
  2. L3:459 suppresses
  3. L19:39 suppresses
  4. L20:111 suppresses
  5. L16:48 excites
  6. L16:14 excites
  7. L16:0 excites
  8. L18:541 excites
  9. L16:27 suppresses
  10. L20:441 excites
  11. L21:195 excites
  12. L18:249 suppresses
  13. L18:569 excites
  14. L2:1083 suppresses
  15. L6:13 excites
  16. L19:231 suppresses
  17. L16:143 excites
  18. L20:198 excites
  19. L8:7 excites
  20. L21:1781 suppresses
  21. L20:1120 weakly excites
  22. L16:107 weakly excites
  23. L16:188 weakly excites
  24. L9:1377 weakly excites

Feeds later layers

  1. L24:2478 excites
  2. L27:1450 excites
  3. L27:2206 excites
  4. L27:325 excites
  5. L26:319 excites
  6. L26:744 weakly suppresses
  7. L27:1687 weakly excites
  8. L27:1494 weakly suppresses
  9. L27:128 weakly excites unclear: opening quotes and closing brackets
  10. L27:2442 weakly suppresses
  11. L26:2930 weakly excites
  12. L23:1506 weakly excites
  13. L24:2026 weakly excites
  14. L27:2504 weakly suppresses
  15. L27:766 weakly excites word endings split mid-word, like 'shower' in 'showered'
  16. L27:775 weakly suppresses
  17. L24:696 weakly suppresses
  18. L24:1988 weakly suppresses
  19. L25:2123 weakly excites
  20. L24:2403 weakly suppresses
  21. L23:1740 weakly excites
  22. L24:2699 weakly suppresses
  23. L24:2338 weakly excites
  24. L23:918 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.

3 values between -3.77 and -3.332 values between -2.45 and -2.014 values between -2.01 and -1.5812 values between -1.58 and -1.1464 values between -1.14 and -0.70371 values between -0.70 and -0.263162 values between -0.26 and 0.18243 values between 0.18 and 0.6226 values between 0.62 and 1.052 values between 1.05 and 1.494 values between 1.49 and 1.932 values between 1.93 and 2.373 values between 2.37 and 2.811 values between 2.81 and 3.251 values between 3.25 and 3.681 values between 3.68 and 4.121 values between 4.12 and 4.561 values between 5.44 and 5.881 values between 5.88 and 6.311 values between 6.75 and 7.191 values between 7.19 and 7.631 values between 10.70 and 11.141 values between 13.33 and 13.77offnegativepositiveoften

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. ing
  2. ING
  3. ğun
  4. eft
  5. 지도
  6. ちは
  7. abd
  8. ドル
  9. 过的
  10. ments

Suppresses

  1. obel
  2. élect
  3. итет
  4. elas
  5. ramer
  6. 不限
  7. ework
  8. 楫
  9. alog
  10. lob