Neuron Explorer

Qwen3-0.6B

'per' in rate expressions

A neuron in layer 10 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

The neuron fires on the word 'per' (or 'each', 'a') when it introduces a rate, like 'people per square mile', '3 inches per year' or '12,000 per family'. It stays mostly quiet elsewhere. Turning it up made the model write unrelated word problems and math-style text, which does not match the trigger.

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

    not want to move to Curry Island with his/her...⏎This section contains 570 words|⏎(approx. 2 pages at 300 words per page)<|endoftext|>Of all the technology featured at CES this week, the most exciting

  2. #2 peak

    there were 112 people, 46 households, and 32 families residing in the town. The population density was 1.2 people per square mile (0.4/km²). There were 106

  3. #3 peak

    7 and located in the area of Asia, with a land area of 796096 km² and population density of 278 people per km². Territory

  4. #4 peak

    875,000 barrels total) with a reasonable profit at the present price of crude oil, and 2,500 barrels per acre for an area of 2,248 acres (5,6

  5. #5 peak

    to one-third in 1 year.⏎Added to these principal items is a reduction in volume per plant. Thus, total manufacturing costs have been increased about a cent a pound, compared with the same months a year ago.⏎Using typical cost figures supplied

  6. #6 peak

    the socioeconomic benefits accrued to translocated farmers," they write. "At an average start-up resettlement cost of at least US$12,000 per family, this is an expensive development program."⏎From a macroeconomic standpoint, this

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: how to prevent a fire at all? I have never been helped by any rescue breathing apparatus. A. It is possible to escape. It is possible to escape. It is possible to escape. All Animals are safe from being harmed by a

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

    11, electricians earned an average of $52,910 per year, according to the Bureau of Labor Statistics. The average hourly pay for an electrician was estimated to be $25.44. The median annual

  2. #2 peak

    Bureau of Labor Statistics, earnings for mediators range from $28,090 to $102,202 per year with median annual wages estimated at $49,490. Hourly

  3. #3 peak

    ians. The lowest average salary for electricians, $39,640 per year, was reported in North Carolina.⏎The Bureau of Labor Statistics estimates that job growth

  4. #4 peak

    @example.org for more information. Updated June 25, 2015. (taken from http://genealogy.about.com/od/surname_meaning/p/boyle.htm ) BOYLE Last Name Meaning & Origin: A variant

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. L9:1106 suppresses
  2. L4:141 excites
  3. L8:1850 suppresses
  4. L8:1249 excites
  5. L9:2105 suppresses
  6. L5:2215 excites
  7. L7:1350 suppresses
  8. L8:276 excites
  9. L7:2787 excites
  10. L5:1145 suppresses
  11. L8:1858 suppresses
  12. L9:515 suppresses
  13. L7:590 suppresses
  14. L5:2095 excites
  15. L6:337 excites
  16. L7:322 weakly excites
  17. L8:1537 weakly suppresses
  18. L8:1143 weakly excites
  19. L6:263 weakly excites
  20. L8:2997 weakly excites
  21. L0:2646 weakly excites
  22. L9:1177 weakly suppresses
  23. L4:2468 weakly excites
  24. L7:2287 weakly suppresses cutting and dividing in instructions

Feeds later layers

  1. L14:2097 excites
  2. L15:2029 excites
  3. L12:2621 excites
  4. L11:1407 excites
  5. L12:3035 excites
  6. L12:1931 excites
  7. L12:1980 excites
  8. L17:1454 excites
  9. L14:2704 excites
  10. L13:2389 excites
  11. L11:1066 weakly suppresses
  12. L12:1800 weakly excites
  13. L12:1471 weakly excites
  14. L17:2617 weakly excites
  15. L14:2830 weakly excites
  16. L12:842 weakly excites
  17. L19:2500 weakly excites
  18. L19:2592 weakly suppresses
  19. L13:2211 weakly excites
  20. L11:1654 weakly suppresses
  21. L20:2076 weakly excites
  22. L12:2169 weakly excites
  23. L11:267 weakly excites
  24. L11:655 weakly suppresses

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.

2 values between -1.39 and -1.2316 values between -0.74 and -0.5869 values between -0.58 and -0.42412 values between -0.42 and -0.261254 values between -0.26 and -0.101661 values between -0.10 and 0.06410 values between 0.06 and 0.2247 values between 0.22 and 0.3813 values between 0.38 and 0.556 values between 0.55 and 0.713 values between 0.71 and 0.872 values between 0.87 and 1.031 values between 1.03 and 1.192 values between 1.19 and 1.352 values between 1.35 and 1.513 values between 1.67 and 1.842 values between 2.32 and 2.481 values between 2.80 and 2.961 values between 4.74 and 4.901 values between 4.90 and 5.06offnegativepositiveoften

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. 每次
  2. latable
  3. 速度
  4. 每一次
  5. lse
  6. FINE
  7. 每
  8. الزمن
  9. każd
  10. 次

Suppresses

  1. NewLabel
  2. amous
  3. üz
  4. 无知
  5. circum
  6. reek
  7. Boeh
  8. tec
  9. Sachs
  10. aine