Neuron Explorer

Qwen3-0.6B

"and" closing a number range

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 fires on the word "and" when it joins two numbers in a range, as in "between 1 and 100" or "between 4 and 5.5." Turning it up did not produce any obvious change tied to numbers or ranges.

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 spider after the agile panther Bagheera in Kipling's book."⏎"Utterly Surreal" Spider⏎Between 2001 and 2008, Meehan and colleagues studied the spider in its

  2. #2 peak

    of globalisation: it is becoming more stratified – with the rapid growth of what they term “the developing middle class” – a group on between $4 and $13 a day. This group has grown from 600 million

  3. #3 peak

    ’s First Line of Defense: The Acid Mantle⏎Our skin has an acid mantle with a pH between 4 and 5.5 that forms a protective barrier on its surface. The acid mantle

  4. #4 peak

    and chief engineer of PNNL's solid oxide fuel cell development program. "However, this research shows that smaller solid oxide fuel cells that generate between 1 and 100 kilowatts of power are a viable

  5. #5 peak

    The importance of mixing traditional and new information technologies in strategies designed to improve democratic governance and reduce poverty;⏎- The importance of<|endoftext|>Between the years 900 and 1100, Chinese painters created visions of landscape that depicted the sub

  6. #6 peak

    transmissions, the duration and power output of AMP's smart meters are well within FCC safe exposure levels. When we tested our meters, our readings were between 1 and 6 microwatts per square centimeter when 2 feet away from the

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 reading about the recent events in the country, and 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

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

    shapes with simple mapping between the eye surface and the projection surface. Using the Fibonacci pattern, which is found repeatedly

  2. #2 peak

    importance of connections between the biological diversity of the Sky Islands and that of the Sierra Madre Occidental of

  3. #3 peak

    to establish relationships between the diverse units of information gathered on actors (individual or collective like corporations), actions,

  4. #4 peak

    reactor fuel). To prevent the chain reaction from moving too rapidly, control mechanisms between the fuel containers absorb extra neutrons. As an extra precaution, nuclear plants are shield

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. L15:2994 weakly suppresses
  2. L9:969 weakly suppresses
  3. L1:717 weakly suppresses
  4. L5:1850 weakly suppresses
  5. L10:1214 weakly suppresses
  6. L9:723 weakly excites
  7. L11:1464 weakly suppresses
  8. L23:2589 weakly suppresses
  9. L1:642 weakly excites
  10. L16:294 weakly excites
  11. L18:461 weakly suppresses
  12. L11:2187 weakly excites
  13. L5:822 weakly suppresses
  14. L7:2510 weakly excites
  15. L12:1017 weakly suppresses
  16. L5:751 weakly suppresses
  17. L6:815 weakly suppresses
  18. L13:2061 weakly suppresses
  19. L8:1627 weakly excites
  20. L16:2884 weakly suppresses
  21. L21:1987 weakly suppresses
  22. L26:2899 weakly suppresses
  23. L25:452 weakly suppresses
  24. L6:2474 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 -1.08 and -0.961 values between -0.84 and -0.723 values between -0.60 and -0.483 values between -0.48 and -0.3624 values between -0.36 and -0.2395 values between -0.23 and -0.111888 values between -0.11 and 0.011691 values between 0.01 and 0.13134 values between 0.13 and 0.2543 values between 0.25 and 0.3710 values between 0.37 and 0.494 values between 0.49 and 0.623 values between 0.62 and 0.741 values between 0.74 and 0.861 values between 0.86 and 0.981 values between 1.22 and 1.342 values between 1.71 and 1.831 values between 1.95 and 2.071 values between 3.41 and 3.531 values between 3.65 and 3.77offnegativepositiveoften

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. rogram
  2. .Navigator
  3. imap
  4. EndPoint
  5. .gridView
  6. DataSet
  7. .GridColumn
  8. 成绩单
  9. ilog
  10. rist

Suppresses

  1. zeich
  2. ủy
  3. تسجيل
  4. ighth
  5. LY
  6. ана
  7. NE
  8. بق
  9. ancel
  10. 姐妹