Neuron Explorer

Qwen3-0.6B

unclear: weak, end of web page

A neuron in layer 12 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 gives only weak signals, mostly where a web page or document ends and a new unrelated one begins. It is hard to see a clear meaning, and the model's written text did not change in an obvious, consistent way when it was adjusted.

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

    of a lighthouse. Think of interesting words to describe the looks and feel of the lighthouse. If you can only think of boring adjectives to describe it<|endoftext|>In a creative writing classroom, when the class discusses and analyzes a work of fiction

  2. #2 peak

    anniversary of his death⏎The article you want to access is behind a paywall. You can purchase this article or subscribe to access all the low countries articles.<|endoftext|>Repeated infections can have various causes, ranging in severity from 'worrying'

  3. #3 peak

    examples of creating markers, setting markers, and moving point to markers:⏎;; Make a new marker that initially does not point anywhere: (setq m1 (make-marker)) ⇒ #<marker in no buffer>⏎m1to point between the

  4. #4 peak

    the Ninevites to preach repentance (Jon. 3: 3-10).⏎Our Lord Jesus Christ, addressing the Scribes and the Pharisees who demanded a sign from Him, said that no sign would be given except

  5. #5 peak

    Coverage⏎- 1st Photos of Mars by Curiosity Rover (Gallery)⏎Copyright 2012 SPACE.com, a TechMediaNetwork company. All rights reserved. This material may not

  6. #6 peak

    :⏎- US English dictionary⏎What do you find interesting about this word or phrase?⏎Comments that don't adhere to our Community Guidelines may be moderated or removed.<|endoftext|>Even the hardy tardigrade might not stand up to climate change⏎T

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: what if we can create a new way to design and implement a solution that is based on software and is built to support complex problems, with a common approach, and a way that is easy to implement, to be maintained, and to be

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

    :|| Web page: http://www.tandfonline.com/VECE20|⏎|Order Information:||Web: http://www.tandfonline.com/pricing/journal/VECE20|⏎When requesting a correction,

  2. #2 peak

    23(2), 1976 pp. 125–135. doi:10.1112/S0025579300008718.

  3. #3 peak

    Krystal McLear, 32, who serves on a committee for the network that runs the U.S. organ transplant system.⏎The United Network for Organ Sharing is debating some new policies to change that. Among the proposals: A checklist for

  4. #4 peak

    Relation to “The Giver”⏎Memory is something that happened in the past and something someone experienced and is in

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. L10:2394 suppresses
  2. L9:2282 excites
  3. L9:310 excites
  4. L9:1660 weakly suppresses
  5. L11:993 weakly excites
  6. L5:1603 weakly suppresses
  7. L4:2165 weakly suppresses
  8. L11:985 weakly suppresses
  9. L5:883 weakly suppresses
  10. L3:1475 weakly suppresses
  11. L2:1181 weakly suppresses
  12. L3:203 weakly excites
  13. L10:935 weakly excites
  14. L11:1044 weakly suppresses
  15. L9:2870 weakly excites dictionary definition wording
  16. L9:894 weakly suppresses
  17. L11:891 weakly suppresses
  18. L2:2360 weakly suppresses
  19. L5:2965 weakly excites
  20. L11:834 weakly excites
  21. L5:73 weakly suppresses
  22. L11:2944 weakly suppresses
  23. L10:36 weakly excites
  24. L8:2824 weakly suppresses

Feeds later layers

  1. L16:978 suppresses
  2. L15:1859 suppresses
  3. L14:2050 weakly suppresses
  4. L15:485 weakly suppresses
  5. L13:1205 weakly excites
  6. L16:1629 weakly suppresses
  7. L16:431 weakly excites
  8. L22:232 weakly excites
  9. L13:2661 weakly excites
  10. L27:1319 weakly excites
  11. L17:1088 weakly excites
  12. L15:2127 weakly excites
  13. L24:82 weakly suppresses
  14. L16:1759 weakly excites
  15. L14:1735 weakly suppresses
  16. L13:2063 weakly excites
  17. L17:1957 weakly excites
  18. L23:1112 weakly suppresses
  19. L13:1088 weakly excites
  20. L15:2519 weakly suppresses
  21. L16:735 weakly suppresses
  22. L21:1625 weakly suppresses
  23. L24:2335 weakly suppresses
  24. L16:729 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.

4 values between -0.52 and -0.492 values between -0.46 and -0.434 values between -0.43 and -0.415 values between -0.41 and -0.3811 values between -0.38 and -0.3512 values between -0.35 and -0.3218 values between -0.32 and -0.2929 values between -0.29 and -0.2744 values between -0.27 and -0.2455 values between -0.24 and -0.2163 values between -0.21 and -0.18102 values between -0.18 and -0.15124 values between -0.15 and -0.13189 values between -0.13 and -0.10184 values between -0.10 and -0.07271 values between -0.07 and -0.04268 values between -0.04 and -0.01349 values between -0.01 and 0.02301 values between 0.02 and 0.04291 values between 0.04 and 0.07268 values between 0.07 and 0.10255 values between 0.10 and 0.13219 values between 0.13 and 0.16185 values between 0.16 and 0.18168 values between 0.18 and 0.21129 values between 0.21 and 0.24101 values between 0.24 and 0.2766 values between 0.27 and 0.3055 values between 0.30 and 0.3238 values between 0.32 and 0.3530 values between 0.35 and 0.3823 values between 0.38 and 0.4116 values between 0.41 and 0.4415 values between 0.44 and 0.465 values between 0.46 and 0.495 values between 0.49 and 0.522 values between 0.52 and 0.551 values between 0.55 and 0.581 values between 0.58 and 0.60offnegativepositiveoften

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. 稳定的
  2. standen
  3. perator
  4. Ratings
  5. tures
  6. elled
  7. 过得
  8. pane
  9. .itemView
  10. favor

Suppresses

  1. reat
  2. 魔王
  3. ree
  4. 創
  5. �
  6. iej
  7. (side
  8. aby
  9. reative
  10. ige