Neuron Explorer

Qwen3-0.6B

the digit 0 in years 2000 to 2009

A neuron in layer 22 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 zero in years like 2005 or 2004 when it comes after '20'. Turning it up made the model write about disasters, events and years, though not in a clearly consistent way.

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

    build upon critical lessons learned from 2008's major hurricanes, Gustav and Ike, as well as other powerful storms, such as 2005's Katrina and Rita and 2004's Ivan.⏎API plays

  2. #2 peak

    utopo Purwo Nugroho, a spokesman for the Indonesian National Disaster Management Agency.⏎The extent of the damage is still being assessed.⏎In 2004, a 9.1-magnitude underwater earthquake struck off the coast of

  3. #3 peak

    <|endoftext|>Computer scientists concerned with the systems security of Direct-Recording Electronic (DRE) voting machines are by now familiar with the Diebold debacle of 2005—when activist Bev Harris discovered and then posted the unprotected source code for

  4. #4 peak

    President Luis Inácio da Silva, a man from the working class, became president in 2003, he called for a program to eliminate hunger completely. But that goal has

  5. #5 peak

    and the Gulf Coast region is home to almost half of the U.S. refining capacity.⏎Upstream (Exploration and Production)⏎During the major 2005 hurricanes, waves were higher and winds were stronger than anticipated in deeper parts of

  6. #6 peak

    pp. 532-535, 2013.<|endoftext|>Following the Indian Ocean tsunami of December 26, 2004, the Indonesian government placed restrictions

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 solution to the solution of the given out of the solution of the solution of the solution... (n step solution) is known, and the solution is the solution of the solutions of the solutions... (n+1 step solution) and

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

    is relevant for Professor Finnegan's class at WSU.⏎What textbook is this problem found in?⏎Our data indicates that this problem or a close variation was asked in Chemistry: An Atoms First Approach - Zumdahl Atoms 1st

  2. #2 peak

    Help in Classification Of Relays⏎Transtutors is the best place to get answers to all your doubts regarding classification of relays with examples. Transtutors has a vast panel of experienced in classification of relays electrical engineering tutors who can explain

  3. #3 peak

    , adaptations and important relationships among organisms (symbiosis, predator-prey dynamics and energy transfer) that do not occur on land.⏎Fundamental Concept 5e.⏎The ocean is three-dimensional, offering vast living space and diverse habitats from the

  4. #4 peak

    paper should meet the following requirements:⏎3-4 pages in length (does not include title and reference page)⏎1-2 outside sources⏎Why Work with CheggAnswers⏎At CheggAnswers.com, we boast of our quick turnaround that is

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. L20:1086 suppresses
  2. L20:639 excites
  3. L20:918 suppresses
  4. L2:170 weakly suppresses
  5. L9:1912 weakly suppresses
  6. L14:463 weakly excites
  7. L0:202 weakly suppresses start of a new document
  8. L3:1156 weakly suppresses
  9. L7:2651 weakly excites textbook and study-help wording
  10. L3:52 weakly suppresses
  11. L6:2538 weakly suppresses
  12. L4:168 weakly suppresses
  13. L7:995 weakly suppresses
  14. L6:740 weakly suppresses
  15. L21:127 weakly excites
  16. L2:2606 weakly excites
  17. L6:1957 weakly suppresses
  18. L3:153 weakly suppresses
  19. L14:951 weakly excites
  20. L10:763 weakly excites
  21. L20:506 weakly excites
  22. L21:2384 weakly suppresses
  23. L11:2106 weakly suppresses
  24. L21:2909 weakly suppresses

Feeds later layers

  1. L23:2904 suppresses
  2. L24:1326 suppresses
  3. L23:1119 weakly excites
  4. L27:1034 weakly suppresses
  5. L25:2336 weakly suppresses
  6. L26:3023 weakly suppresses
  7. L23:2993 weakly suppresses
  8. L25:983 weakly excites
  9. L24:1868 weakly suppresses
  10. L24:1898 weakly excites
  11. L24:222 weakly suppresses
  12. L23:2866 weakly suppresses
  13. L27:1526 weakly suppresses
  14. L25:2820 weakly excites
  15. L23:542 weakly suppresses
  16. L27:2113 weakly suppresses
  17. L25:1816 weakly excites
  18. L27:1248 weakly suppresses
  19. L24:2641 weakly suppresses
  20. L25:2266 weakly suppresses
  21. L24:578 weakly suppresses
  22. L24:1713 weakly suppresses
  23. L23:2290 weakly suppresses
  24. L24:2811 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 -23.98 and -23.281 values between -11.94 and -11.233 values between -6.27 and -5.563 values between -5.56 and -4.851 values between -4.85 and -4.144 values between -4.14 and -3.435 values between -3.43 and -2.724 values between -2.72 and -2.026 values between -2.02 and -1.3160 values between -1.31 and -0.603110 values between -0.60 and 0.11682 values between 0.11 and 0.8220 values between 0.82 and 1.536 values between 1.53 and 2.241 values between 2.24 and 2.951 values between 3.66 and 4.36offnegativepositiveoften

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. anner
  2. achelor
  3. 髭
  4. bable
  5. gorithms
  6. olicies
  7. >About
  8. quier
  9. umes
  10. GORITH

Suppresses

  1. solutions
  2. solution
  3. (solution
  4. Solutions
  5. olution
  6. solution
  7. olutions
  8. _solution
  9. Solution
  10. 答案