Neuron Explorer

Qwen3-0.6B

unclear: stray list markers and periods

A neuron in layer 0 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 is weak and hard to pin down. It fires a little on list numbers like "1" in an opening bracket and more strongly goes quiet, or negative, on full stops that start a new line. Turning it up did not produce a clear story.

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

    land⏎(1808–1867) Hannah was born and raised in a

  2. #2 peak

    terminated:⏎(1) The battery surface temperature drops 20% lower than the peak value;⏎

  3. #3 peak

    bonded warehouse.⏎By cargo characteristics⏎(1) Raw material warehouse: The raw material warehouse is used to store the raw materials

  4. #4 peak

    ⏎1. n. Any one of numerous species of marine gastropods belonging to N

  5. #5 peak

    40% lean protein, 20% fat and 40%

  6. #6 peak

    1:x in x.⏎For non-strict semantics it would be ok to store

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 following things: the world, the sky, the moon, the stars, the planets, the planets, the sun, the earth, the moon, the stars, the planets, the moon, the stars,

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

    the Brotherhood included:⏎The list is instructive because it shows how quickly the influence of the Brotherhood spread through international photographic exhibitions modelled on the annual London Salon⏎. The membership of the Linked Ring⏎soon included photographers from Austria (Hugo Hen

  2. #2 peak

    direct the report to the appropriate office for investigation.<|endoftext|>Skip to main content⏎Create interactive lessons using any digital content including wikis with our free sister product⏎. Get it on the⏎Pages and Files⏎Democratic Teaching Methods⏎Hidden Histories

  3. #3 peak

    ⏎movies are the standard for entertainment. english⏎literature gains popularity across the world. The computer and internet are more accessible through knowledge of english⏎. Professional standards and manuals are first published in english⏎and then translated to other languages

  4. #4 peak

    not included as part of eBook file.<|endoftext|>is a former municipality in Akershus⏎, which lends its name to a municipality and a county in Norway⏎. The name originally belonged to a farm which was located near the current Old Aker

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

None: this is the first layer.

Feeds later layers

  1. L4:676 suppresses
  2. L2:280 suppresses
  3. L3:378 suppresses
  4. L3:2426 weakly suppresses
  5. L9:2764 weakly suppresses
  6. L14:338 weakly excites
  7. L22:314 weakly excites
  8. L11:2106 weakly excites
  9. L24:609 weakly suppresses
  10. L21:1208 weakly suppresses
  11. L21:253 weakly suppresses
  12. L14:2658 weakly excites
  13. L14:289 weakly suppresses
  14. L5:2936 weakly excites
  15. L25:1017 weakly excites
  16. L3:2278 weakly suppresses
  17. L24:2940 weakly suppresses
  18. L24:2755 weakly suppresses
  19. L2:814 weakly suppresses
  20. L6:2262 weakly suppresses
  21. L10:2608 weakly suppresses
  22. L25:39 weakly excites
  23. L21:441 weakly excites
  24. L7:756 weakly 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 -1.68 and -1.622 values between -1.00 and -0.951 values between -0.44 and -0.382 values between -0.33 and -0.272 values between -0.27 and -0.218 values between -0.21 and -0.1699 values between -0.16 and -0.10700 values between -0.10 and -0.041602 values between -0.04 and 0.01910 values between 0.01 and 0.07207 values between 0.07 and 0.1345 values between 0.13 and 0.1817 values between 0.18 and 0.2466 values between 0.24 and 0.3080 values between 0.30 and 0.35142 values between 0.35 and 0.4121 values between 0.41 and 0.471 values between 0.47 and 0.522 values between 0.52 and 0.58offnegativepositiveoften

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. ud
  2. ude
  3. ew
  4. UMENT
  5. ush
  6. 卑
  7. umer
  8. mentation
  9. UM
  10. pletion

Suppresses

  1. rido
  2. ResponseBody
  3. 涨停
  4. accion
  5. yect
  6. ored
  7. arrêt
  8. erin
  9. ʝ
  10. 点点头