Neuron Explorer

Qwen3-0.6B

'from' after change or increase verbs

A neuron in layer 2 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 "from" when it marks a starting point after words about change, like "increased from 74 percent" or "changed from light orange". Turning it up did not clearly produce text about change.

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

    05 were identified as achieving above the national reading benchmarks and 91.6 percent in writing.⏎The percentage of staff involved in professional development has increased from 74 percent to 88 percent.⏎Thornland State School '

  2. #2 peak

    per year. This growth leads to the growth of the transit fees paid by and paid to the upstream ISPs.⏎For example, while the price for Internet Transit drops from $3/Mbps down to $2/Mbps, the volume of traffic during

  3. #3 peak

    prevalence of helmet wearing among skiers from 1996 to 2012. They noted that the number of skiers wearing helmets had increased from 8% to 84%, and that the number of reported head injuries

  4. #4 peak

    living substance emits its own electromagnetic frequencies. Our understanding of the human body is shifting from a mechanical-chemical perspective to a quantum-energetic perspective.⏎Dr. Robert

  5. #5 peak

    , making it easier for them to swim under Arctic ice sheets.⏎In the summer, these small whales migrate up and down the river each day as the tide changes from low to high. They travel in large groups called pods. When winter approaches they

  6. #6 peak

    1947 and 1948 have increased from 7 percent to 15 percent over those in 1946

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: ...and what if you copy this. The following is a very common question – I'll use x, and remember to answer my own questions. Okay, I'm sorry for any of you, but I'm not sure if I'm being

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

    Clause in oracle. In following section i would like to explain about the syntax as well as multiple examples of with clause.⏎With SQL_Query_Name As⏎Select * from New_SQL_Query_name;⏎Execution of With Clause :⏎In this section i would like

  2. #2 peak

    stage of disease is characterized by fasting hyperglycemia, glucosuria and usually three p’s that are polyuria, polydipsia and polyphagia.⏎

  3. #3 peak

    r, pioneer cultural anthropologists Franz Boas and Margaret Mead, and ornithologist Robert Cushman Murphy. J. P. Morgan was among famous benefactors of the Museum.⏎American Museum of Natural History, Central Park West at 7

  4. #4 peak

    is often diagnosed during a visit to your doctor. Blood pressure is measured using a cuff around your arm and a device called a sphygmomanometer. Your doctor may ask you to sit quietly for five minutes before checking

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. L0:1660 weakly suppresses
  2. L1:1105 weakly excites
  3. L0:709 weakly excites
  4. L1:2987 weakly excites
  5. L1:1583 weakly excites
  6. L1:835 weakly excites
  7. L1:1222 weakly suppresses
  8. L0:1565 weakly excites
  9. L0:1380 weakly excites
  10. L1:934 weakly excites
  11. L0:2623 weakly excites
  12. L0:178 weakly suppresses
  13. L0:248 weakly suppresses
  14. L1:371 weakly excites
  15. L0:1301 weakly suppresses
  16. L0:1717 weakly suppresses
  17. L0:1174 weakly suppresses
  18. L1:1592 weakly excites
  19. L1:2116 weakly excites
  20. L1:922 weakly excites
  21. L1:616 weakly suppresses
  22. L0:2341 weakly suppresses
  23. L1:14 weakly excites
  24. L1:2119 weakly excites

Feeds later layers

  1. L19:777 weakly excites
  2. L7:2715 weakly excites
  3. L7:2808 weakly suppresses
  4. L18:1108 weakly excites
  5. L5:1976 weakly suppresses
  6. L6:2447 weakly excites
  7. L5:1605 weakly suppresses
  8. L18:2225 weakly excites
  9. L20:697 weakly excites
  10. L7:2322 weakly suppresses
  11. L21:2011 weakly excites
  12. L17:2049 weakly excites
  13. L4:1954 weakly suppresses
  14. L7:856 weakly suppresses
  15. L9:1343 weakly excites
  16. L4:2468 weakly suppresses
  17. L13:1750 weakly excites
  18. L26:1999 weakly suppresses
  19. L26:1875 weakly excites
  20. L13:1211 weakly excites
  21. L13:1097 weakly suppresses
  22. L21:183 weakly suppresses
  23. L15:1127 weakly excites
  24. L23:92 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.

1 values between -1.38 and -1.292 values between -1.20 and -1.111 values between -0.93 and -0.842 values between -0.66 and -0.572 values between -0.57 and -0.482 values between -0.48 and -0.3915 values between -0.39 and -0.3077 values between -0.30 and -0.21303 values between -0.21 and -0.121035 values between -0.12 and -0.031504 values between -0.03 and 0.06714 values between 0.06 and 0.15186 values between 0.15 and 0.2441 values between 0.24 and 0.335 values between 0.33 and 0.422 values between 0.42 and 0.512 values between 0.51 and 0.602 values between 0.60 and 0.691 values between 0.69 and 0.782 values between 0.78 and 0.871 values between 0.87 and 0.963 values between 0.96 and 1.051 values between 1.50 and 1.592 values between 1.68 and 1.771 values between 2.04 and 2.131 values between 2.13 and 2.22offnegativepositiveoften

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. tdown
  2. ificaciones
  3. versions
  4. daughter
  5. сан
  6. ��
  7. %%⏎
  8. ?><?
  9. 最低
  10. 在床上

Suppresses

  1. sters
  2. oku
  3. .Commit
  4. ster
  5. 睛
  6. bức
  7. estre
  8. guide
  9. 时刻
  10. 帮忙