Neuron Explorer

Qwen3-0.6B

web citations, urls and dateline dashes

A neuron in layer 14 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 inside web addresses, reference lists and article credits, such as the dash in a link or the start of a citation. It is strongly pushed negative by the dash after a city name at the start of a news story. Turning it up gave loose, rambling text that did not clearly follow this pattern.

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

    Goods and Deforestation. voluntary private sector purchasing policies and investment standards. However, there has been limited analysis⏎Letter To Editor Regarding The Concern For Deforestation. Dear Editor, Monika’s Doggie Rescue covered a holistic statistics released by NSW government

  2. #2 peak

    scientists with a strong interest in teaching.⏎Folkert van Werven, and Alexi Goranov. 7.342 To Divide or Not To Divide: Control of Cell Cycle and Growth by Extracellular Cues, Fall

  3. #3 peak

    full article here: http://ourgame.mlblogs.com/2014/08/18/out-at-home/⏎- Related link: Click here to view photos from the 17

  4. #4 peak

    ,” he said.⏎For the rest of this article, click here: http://www.abc.net.au/news/rural/2017-05-23/the-future-of-automation-in-the-mining-industry/8

  5. #5 peak

    : Philosophy and the New Science in the University. The Hague,Nijhoff.⏎Gordon Fraser (ed.) (2009). The New Physics for the Twenty-First Century. Cambridge University Press.⏎C. W. Kilmister

  6. #6 peak

    to the full article PDF.⏎Price includes VAT (USA)⏎Tax calculation will be finalised during checkout.⏎Feng ZW, Cao HF, Zhou XP (1999) The effects of acid precipitation on the ecological environment and restoration

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 world's most powerful nation-state in the Earth's landlocked country, the foreign policy officials in the West have confirmed that the government's foreign policy team has been on the ground in the southern Africa region in the western Africa coast south of

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

    à jour : 05/07/2021⏎Ajouter un commentaire<|endoftext|>By DANIEL SUDDEATH⏎NEW ALBANY —⏎Judging by the pleasure clearly shown by the children as they wheeled their

  2. #2 peak

    the body, and consequently a bend is formed at the<|endoftext|>DENVER – Put on your poodle skirts and tune in Elvis on the transistor radio, because

  3. #3 peak

    aP2 acts as a signal in cells, setting off a chain of inflammatory and<|endoftext|>, January 13, 2010 (ENS) - Eight landfill methane capture projects were recognized by the U.S. EPA Tuesday for innovation

  4. #4 peak

    . Your student might just be flattered (if not shocked) that you'd be insightful enough to ask.⏎By Rod Machado<|endoftext|>PARIS — Should Palmyra be rebuilt? And under what conditions? No sooner had Syrian

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. L13:1221 strongly suppresses comma or verb after a researcher or study subject
  2. L11:1391 strongly excites
  3. L12:978 strongly suppresses
  4. L13:1869 excites
  5. L13:1115 excites
  6. L13:1152 suppresses
  7. L13:257 excites
  8. L10:2212 suppresses
  9. L7:741 suppresses
  10. L13:52 excites
  11. L12:1316 excites
  12. L12:1176 suppresses
  13. L13:1501 suppresses
  14. L10:2988 suppresses
  15. L7:1766 suppresses
  16. L7:1454 excites
  17. L13:1414 excites
  18. L11:1974 suppresses
  19. L13:1558 suppresses
  20. L6:1449 excites
  21. L11:595 suppresses
  22. L8:1072 suppresses
  23. L7:3050 excites
  24. L12:2758 excites

Feeds later layers

  1. L15:1687 suppresses
  2. L18:796 suppresses
  3. L16:1903 suppresses
  4. L17:808 excites
  5. L15:2008 suppresses
  6. L18:2980 suppresses
  7. L15:2134 suppresses
  8. L16:2895 suppresses
  9. L19:768 suppresses
  10. L16:2772 suppresses
  11. L18:1031 suppresses
  12. L20:2429 excites
  13. L15:1046 weakly excites
  14. L16:1688 weakly excites
  15. L17:2281 weakly excites
  16. L16:2007 weakly suppresses
  17. L17:2032 weakly excites
  18. L15:332 weakly suppresses
  19. L17:2320 weakly excites
  20. L15:1504 weakly suppresses
  21. L20:2495 weakly suppresses
  22. L15:290 weakly suppresses
  23. L16:1887 weakly excites
  24. L17:2329 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 -4.52 and -4.371 values between -3.31 and -3.161 values between -3.01 and -2.862 values between -2.86 and -2.712 values between -2.71 and -2.563 values between -2.56 and -2.411 values between -2.41 and -2.263 values between -2.26 and -2.111 values between -2.11 and -1.963 values between -1.96 and -1.816 values between -1.81 and -1.664 values between -1.66 and -1.519 values between -1.51 and -1.3612 values between -1.36 and -1.2112 values between -1.21 and -1.0621 values between -1.06 and -0.9124 values between -0.91 and -0.7620 values between -0.76 and -0.6145 values between -0.61 and -0.4557 values between -0.45 and -0.30134 values between -0.30 and -0.15662 values between -0.15 and -0.001383 values between -0.00 and 0.15971 values between 0.15 and 0.30399 values between 0.30 and 0.45100 values between 0.45 and 0.6022 values between 0.60 and 0.754 values between 0.75 and 0.902 values between 0.90 and 1.052 values between 1.05 and 1.201 values between 1.35 and 1.50offnegativepositiveoften

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. afterward
  2. /logout
  3. :</
  4. 班车
  5. UnderTest
  6. iqué
  7. @endif
  8. →then
  9. then
  10. *)&

Suppresses

  1. 的一项
  2. 有一种
  3. 悄然
  4. 近日
  5. สำรวจ
  6. 的世界
  7. OURNAL
  8. vue
  9. unthinkable
  10. vis