Neuron Explorer

Qwen3-0.6B

last digit of years like 1970

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 lights up on the final zero in numbers like 1970, 1970s or 2500, especially years in the 1970s. Turning it up made the model's writing break into repetitive loops, so the steering text tells us little about it.

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

    the overall picture of the conflict between two worlds, East and West, which determined the course of history.⏎Proving History is available online: https://2500years.culture.gov.gr/en.<|endoftext|>"Tell all the Truth"⏎Tell

  2. #2 peak

    thought.⏎The results are to be published in the journal Science.⏎The lower mantle comprises 55 percent of the planet by volume and extends from 670 and 2900 kilometers in depth, as defined by the so-called

  3. #3 peak

    umatic road tube traffic counter. Rubber hoses are stretched across a road and passing vehicles create air pulses that are recorded by a roadside counter. In 1970 the counts were mechanically recorded on a roll of paper tape. The time and number

  4. #4 peak

    are considered the progenitors of the modern Ukrainian nation and Ukrainian nationalist discourse appeals, to a large extent, to the Cossacks.⏎In 1670, the Don Cossack Stenka Razin proclaimed<|endoftext|>Amateur astronomers

  5. #5 peak

    landscapes beyond protected area networks.⏎“Remote sensing data from orbiting satellites have been used to measure, understand, and predict environmental changes since the 1970s, but technology that subsequently became available can now be applied much more widely on

  6. #6 peak

    create transition matrices of land cover change. Dates of acquired images represent the most cloud-free image retrievals from 1970-2001 for each site and are therefore not continuous. There are

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 lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately

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

    and%20adaptive%20immune%20system.⏎- How Vitamin C Supports a Healthy Immune Sysstem. Academy of Nutrition

  2. #2 peak

    be 99,800 unaffected pregnancies and 200 pregnancies with Down syndrome. The table below compares the

  3. #3 peak

    Mentor: Carl Ruetz⏎Page last modified July 14, 2009<|endoftext|>what does hamlet think about suicide? with textual support⏎2 Answers | Add Yours

  4. #4 peak

    called Wheel. If we consider the real-world, we can find many objects around us,<|endoftext|>18th Jan 12 14:01<< Go back to news⏎The kinetic force of braking can now be used in hybrid

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. L1:82 suppresses
  2. L1:1458 suppresses
  3. L5:263 excites
  4. L3:230 suppresses
  5. L13:1160 suppresses
  6. L3:856 weakly suppresses
  7. L16:1220 weakly suppresses
  8. L15:199 weakly excites
  9. L8:2711 weakly suppresses
  10. L15:623 weakly excites
  11. L6:227 weakly suppresses
  12. L2:398 weakly suppresses
  13. L5:1219 weakly suppresses
  14. L1:2557 weakly suppresses
  15. L2:1774 weakly suppresses
  16. L3:282 weakly excites
  17. L1:1646 weakly suppresses
  18. L15:2065 weakly suppresses
  19. L25:2606 weakly suppresses
  20. L27:1343 weakly suppresses
  21. L16:2621 weakly suppresses
  22. L6:2970 weakly excites
  23. L25:2814 weakly suppresses
  24. L18:1308 weakly suppresses opening parenthesis before an acronym

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.

3 values between -0.61 and -0.5225 values between -0.52 and -0.4333 values between -0.43 and -0.3410 values between -0.34 and -0.2512 values between -0.25 and -0.16322 values between -0.16 and -0.072665 values between -0.07 and 0.02716 values between 0.02 and 0.1240 values between 0.12 and 0.215 values between 0.30 and 0.396 values between 0.39 and 0.486 values between 0.48 and 0.572 values between 0.57 and 0.666 values between 0.66 and 0.752 values between 0.84 and 0.932 values between 0.93 and 1.025 values between 1.02 and 1.122 values between 1.12 and 1.2110 values between 1.21 and 1.306 values between 1.30 and 1.393 values between 1.39 and 1.483 values between 1.48 and 1.572 values between 1.57 and 1.662 values between 1.66 and 1.755 values between 1.75 and 1.843 values between 1.84 and 1.935 values between 1.93 and 2.021 values between 2.02 and 2.121 values between 2.12 and 2.211 values between 2.66 and 2.752 values between 2.75 and 2.842 values between 2.93 and 3.02offnegativepositiveoften

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. hma
  2. 地中
  3. via
  4. 年后
  5. 禁
  6. th
  7. 送
  8. ustomer
  9. ana
  10. bj

Suppresses

  1. abble
  2. ITS
  3. endregion
  4. fection
  5. /owl
  6. ール
  7. 六大
  8. fections
  9. реш
  10. abouts