Neuron Explorer

Qwen3-0.6B

the 9 in years starting 19

A neuron in layer 23 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 fires on the second digit of years that begin with 19, such as the 9 in 1950s or 1971. Turning it up did not change the topic of what the model wrote, so its effect on the text is hard to see.

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

    monaut. Flew on Soyuz 7, Soyuz 16. More...⏎Grechko Grechko, Georgi Mikhailovich (1931-) Russian engineer cosmonaut. Flew on Soyuz 1

  2. #2 peak

    atta Facts⏎Mohammad Hatta (1902-1980), one of the foremost intellectuals in the Afro

  3. #3 peak

    employment rates for African Americans have been far higher than those of whites for the past 50 years, even in good times. In fact, since 1960 the black unemployment rate has been about twice the white rate. Had blacks

  4. #4 peak

    was to build a device to read the traffic tapes directly and eliminate the tedious manual work. The Intel 8008 microprocessor was announced in 19

  5. #5 peak

    976 at 16.8 cigarettes a day.⏎20% of men currently smoke – the lowest value since the series began in the 1970s. In 1974, half (51%)

  6. #6 peak

    8 election to hard-line nationalists who institutionalised apartheid. He continued to work for reconciliation and emphasised the British Commonwealth's positive role until his death in 19

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: how to use a Ruffle Sware in a Mufsa Muzari. Mufa Muzari is an auster and auster, and the Muzari is a part of the Tariq and Muf

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

    22:6⏎“Those things, which ye have both learned, and received, and heard, and seen in me, do: and the God of peace will be with you”

  2. #2 peak

    479 BC), a Chinese scholar, politician and philosopher who lived in the Spring and Autumn period. The main body of the Confucian canon comprises the Four Books and the Five Classics (四書五經), texts which have been traditionally

  3. #3 peak

    2008 IDE⏎- Create an empty project by select File >> New >> Project. Select General >> Empty Project and type in the project name.⏎- Under folder Source, right-click Select Add >> New Item.⏎- Under Code, Select

  4. #4 peak

    the water. If the water stays cold then you could keep the chicken there a long time, however, once the water is above 40 F (4.4 C) then limit the time.⏎Should I soak chicken in salt water?⏎

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. L21:153 suppresses
  2. L22:1855 excites
  3. L19:417 suppresses
  4. L21:582 excites
  5. L22:399 excites
  6. L20:2181 suppresses
  7. L18:2361 excites
  8. L21:2003 suppresses
  9. L21:475 suppresses
  10. L22:1005 suppresses
  11. L18:2926 weakly suppresses
  12. L2:1321 weakly excites
  13. L17:1336 weakly excites
  14. L16:1036 weakly suppresses
  15. L18:569 weakly suppresses
  16. L10:482 weakly excites
  17. L19:2605 weakly suppresses
  18. L21:1329 weakly suppresses
  19. L20:355 weakly excites
  20. L21:295 weakly excites
  21. L18:576 weakly excites
  22. L17:744 weakly excites
  23. L17:1037 weakly suppresses
  24. L18:2656 weakly suppresses

Feeds later layers

  1. L26:1009 suppresses
  2. L26:745 suppresses
  3. L24:747 suppresses
  4. L27:331 suppresses
  5. L24:340 suppresses
  6. L25:735 excites
  7. L27:870 weakly suppresses
  8. L27:1107 weakly suppresses
  9. L26:132 weakly suppresses
  10. L24:174 weakly excites
  11. L24:105 weakly suppresses
  12. L26:116 weakly suppresses
  13. L24:512 weakly excites 'the' inside organisation and convention names
  14. L25:1623 weakly excites
  15. L25:414 weakly suppresses
  16. L27:1347 weakly excites
  17. L24:2460 weakly suppresses
  18. L25:812 weakly excites
  19. L26:934 weakly suppresses
  20. L27:2059 weakly excites
  21. L27:851 weakly suppresses first names of famous people
  22. L24:27 weakly suppresses
  23. L27:470 weakly suppresses
  24. L25:1487 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.

1 values between -12.76 and -11.792 values between -9.85 and -8.894 values between -8.89 and -7.926 values between -7.92 and -6.9515 values between -6.95 and -5.9833 values between -5.98 and -5.0256 values between -5.02 and -4.05115 values between -4.05 and -3.08169 values between -3.08 and -2.11317 values between -2.11 and -1.15453 values between -1.15 and -0.18619 values between -0.18 and 0.79629 values between 0.79 and 1.76482 values between 1.76 and 2.73397 values between 2.73 and 3.69227 values between 3.69 and 4.66151 values between 4.66 and 5.63110 values between 5.63 and 6.6048 values between 6.60 and 7.5723 values between 7.57 and 8.5318 values between 8.53 and 9.5011 values between 9.50 and 10.476 values between 10.47 and 11.447 values between 11.44 and 12.403 values between 12.40 and 13.371 values between 13.37 and 14.341 values between 17.24 and 18.211 values between 18.21 and 19.181 values between 20.15 and 21.111 values between 21.11 and 22.081 values between 24.99 and 25.95offnegativepositiveoften

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. BJP
  2. CBD
  3. IClient
  4. apist
  5. SPDX
  6. Jedi
  7. WWE
  8. HuffPost
  9. iPod
  10. opause

Suppresses

  1. 萨
  2. チェ
  3. 佩
  4. 迈
  5. 查
  6. Mon
  7. 达
  8. Char
  9. Chip
  10. 芒