Neuron Explorer

Qwen3-0.6B

'at' in 'at least' and 'most' or 'least' qualifiers

A neuron in layer 24 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 'at' in phrases like 'at least' or 'at most', and on similar 'most' and 'least' words. When turned up, the model got stuck repeating 'most' and 'least', which matches the kind of word it likes.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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

    this planet?⏎Or does it lend credence to the idea that life could spring from inanimate material on this planet by showing that it happened not once, but at least twice?⏎Either way, expect the religious fundies to shrug it off

  2. #2 peak

    in freely moving animals. It is also capable of long<|endoftext|>According to legend, the first Valentine greeting was sent by St. Valentine himself. Actually, there are at least three

  3. #3 peak

    Vacciniums I think I may have identified here goes by the common names of Black Highbush Blueberry and Downy Swamp-Huckleberry. There are at least 3 species of Vaccinium/Gaylussacia in the photo above

  4. #4 peak

    A similar outflow is probably present under the disk as well and is hinted at in this illustration. The total amount of material entrained in the broad flow is at least 400 times the mass of the Sun per year. Note that

  5. #5 peak

    expression consists of a single line containing a positive integer number, an operator (one of +, - and *) and the second positive integer number. Every number has at most 500 digits. There are no spaces on the line. If

  6. #6 peak

    is underway in Australia to isolate exactly the type of tick and the specific bacteria that is causing Lyme disease here (there are at least 14 different types of bacteria that cause Lyme disease identified world wide).⏎

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 small 1666-1667. In the small 1666-1667, the 1666-1667 and the 1666-16

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

    15 different kinds of cancer are linked to smoking and other tobacco use. Nearly one-third of all cancer deaths are due to smoking. Quitting smoking is one of the best things you can do for yourself and for those you love. Click

  2. #2 peak

    choices that can reduce risk, including:⏎- Being physically active – More and more research indicates that physical exercise is one of the most important lifestyle choices for reducing the risk of developing Alzheimer’s. Exercise

  3. #3 peak

    talented instructors at Veritas Prep teach students how to prepare for GRE Verbal Reasoning questions. There are several practical strategies available to students that can help them conquer even the most difficult questions in this section.⏎Check out some valuable tips that students may

  4. #4 peak

    most snakes?⏎Florida is considered to be the state with the most snake species in the US, but is it also<|endoftext|>Be a Dent

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. L21:1741 excites
  2. L20:361 suppresses
  3. L21:614 suppresses
  4. L17:30 excites
  5. L18:1547 excites
  6. L19:427 excites
  7. L11:331 suppresses
  8. L18:1084 excites
  9. L16:419 suppresses
  10. L19:644 weakly suppresses
  11. L4:49 weakly excites
  12. L1:59 weakly suppresses 'the' after a verb or preposition
  13. L21:1072 weakly suppresses
  14. L20:219 weakly suppresses "the" after time or place prepositions
  15. L19:2115 weakly suppresses
  16. L9:59 weakly suppresses "a" meaning per (a day, a year)
  17. L6:328 weakly suppresses
  18. L18:861 weakly excites
  19. L15:2587 weakly excites
  20. L18:641 weakly excites
  21. L17:2129 weakly excites
  22. L11:2473 weakly suppresses
  23. L11:875 weakly excites
  24. L9:24 weakly suppresses "the" before a named organisation or title

Feeds later layers

  1. L27:1338 strongly excites
  2. L26:1525 strongly excites
  3. L27:1086 strongly excites
  4. L27:2450 excites
  5. L25:2905 excites
  6. L26:631 suppresses
  7. L25:2986 weakly excites '1' in 21st or 11th, before 'st'
  8. L27:1060 weakly excites
  9. L26:670 weakly suppresses
  10. L26:2537 weakly excites
  11. L27:1811 weakly suppresses
  12. L26:2398 weakly excites
  13. L26:2458 weakly suppresses
  14. L25:2844 weakly excites
  15. L26:1094 weakly excites
  16. L25:626 weakly suppresses
  17. L27:2871 weakly suppresses
  18. L25:2511 weakly suppresses
  19. L27:1296 weakly excites
  20. L25:2102 weakly suppresses
  21. L26:1941 weakly excites
  22. L25:2221 weakly excites
  23. L27:662 weakly excites
  24. L25:1407 weakly suppresses 'in' right after a place or time

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 -36.13 and -34.601 values between -31.55 and -30.031 values between -25.46 and -23.932 values between -23.93 and -22.411 values between -19.36 and -17.833 values between -17.83 and -16.315 values between -16.31 and -14.797 values between -14.79 and -13.2615 values between -13.26 and -11.7415 values between -11.74 and -10.2133 values between -10.21 and -8.6954 values between -8.69 and -7.17102 values between -7.17 and -5.64193 values between -5.64 and -4.12276 values between -4.12 and -2.59542 values between -2.59 and -1.072125 values between -1.07 and 0.46413 values between 0.46 and 1.9857 values between 1.98 and 3.5131 values between 3.51 and 5.0312 values between 5.03 and 6.558 values between 6.55 and 8.083 values between 8.08 and 9.603 values between 9.60 and 11.132 values between 12.65 and 14.171 values between 14.17 and 15.702 values between 23.32 and 24.84offnegativepositiveoften

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. 最
  2. best
  3. least
  4. most
  5. 最好的
  6. least
  7. 最佳
  8. most
  9. 最有
  10. 最大的

Suppresses

  1. nton
  2. uden
  3. vt
  4. tat
  5. rán
  6. 俟
  7. ведь
  8. vation
  9. vertiser
  10. ��