Neuron Explorer

Qwen3-0.6B

sentence-ending periods in explanations

A neuron in layer 12 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 is most active on the full stop that closes a sentence in plain explanatory writing, such as advice or a factual claim. It is least active on full stops after descriptive facts about animals or people. Turning it up did not change the topic in an obvious way.

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

    take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.<|endoftext|>- Major Systems >⏎- The Plumbing Rough-In⏎The Plumbing Rough-In⏎

  2. #2 peak

    don’t know the return column. For example, depending on the specifications, you want to return the value from either column B, C or D and so on. In such case, we are speaking of 2D lookups or a two

  3. #3 peak

    words can also be considered Communized synonyms. In case you want even more details, you can also consider checking out all of the definitions of the word Communized. If there is a match we also include idioms & quotations that either use this

  4. #4 peak

    tmeter. Measure total (supply) voltage with the same voltmeter. You will discover that the voltage drops do not add up to equal the total voltage. This is due to phase shifts in the circuit: voltage dropped across the capacitors

  5. #5 peak

    becausetoo much light from screens before bedtime confuses your body. In fact, it gives it the impression that you aren’t ready to doze off. Blue light from your phone or laptop deceives your brain, as it will think

  6. #6 peak

    . Also take the internal temperatures of all pieces of the chicken. For example, if you are baking six breast pieces, each one might cook differently based on size. Don’t forget to clean your thermometer probe between readings to avoid cross-contamination.⏎

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 to be able to see the color of the color. I am trying to understand what I have to do with the colors, and I have to be able to see colors. I have to be able to understand the difference between the

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

    Grizzly Bear⏎The Grizzly Bear is a type of brown bear. They actually come in a wide variety of colors ranging from almost white to a dark brown. One of their more distinguishing features is the hump on their shoulders, which is

  2. #2 peak

    as Threatened or Endangered⏎Life HistoryThe long-legged myotis is described as having cinnamon red to dark brown fur above and lighter brown to buff below. Hair extends from the body outward to the elbow and knee on the wing and on

  3. #3 peak

    born on 24 May 1870, at the family farm, Bovenplaats, near Malmesbury, in the Cape Colony. His parents, Jacob<|endoftext|>The main purpose of the script is to compare maximum land

  4. #4 peak

    born on 23 January, 1897 in Cuttack (Orissa) to Janakinath Bose and Prabhavati Devi. Janakinath Bose was one of the successful lawyer in Cuttack and received

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. L10:478 suppresses
  2. L9:458 excites
  3. L9:16 suppresses unclear: periods inside numbers and names
  4. L11:132 weakly excites period ending an opening topic sentence
  5. L10:1392 weakly suppresses
  6. L11:1602 weakly suppresses
  7. L6:2962 weakly suppresses
  8. L2:1265 weakly suppresses 'of' in 'this type of'
  9. L5:2804 weakly excites
  10. L4:356 weakly excites
  11. L2:1942 weakly excites
  12. L1:2963 weakly excites
  13. L10:612 weakly excites full stop ending boilerplate or a title
  14. L9:671 weakly suppresses
  15. L10:2944 weakly excites
  16. L1:308 weakly suppresses
  17. L4:2779 weakly suppresses
  18. L7:969 weakly suppresses
  19. L2:606 weakly suppresses
  20. L10:144 weakly suppresses
  21. L7:1613 weakly excites
  22. L10:454 weakly suppresses
  23. L1:1698 weakly excites
  24. L9:12 weakly suppresses

Feeds later layers

  1. L13:517 suppresses comma closing a list item
  2. L20:560 weakly excites
  3. L19:883 weakly suppresses
  4. L15:1948 weakly suppresses
  5. L14:1502 weakly suppresses
  6. L15:27 weakly suppresses
  7. L17:43 weakly suppresses
  8. L25:37 weakly suppresses
  9. L26:373 weakly excites
  10. L16:1097 weakly suppresses
  11. L25:2673 weakly excites
  12. L16:2279 weakly suppresses
  13. L13:403 weakly suppresses
  14. L13:519 weakly suppresses
  15. L21:1892 weakly excites
  16. L23:1195 weakly suppresses
  17. L14:2474 weakly excites
  18. L13:2687 weakly suppresses
  19. L15:1504 weakly excites
  20. L13:101 weakly suppresses
  21. L19:3018 weakly excites
  22. L26:2098 weakly suppresses
  23. L21:2079 weakly excites
  24. L27:346 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 -2.18 and -2.102 values between -1.84 and -1.763 values between -1.76 and -1.673 values between -1.67 and -1.584 values between -1.58 and -1.504 values between -1.50 and -1.412 values between -1.41 and -1.334 values between -1.33 and -1.244 values between -1.24 and -1.162 values between -1.16 and -1.076 values between -1.07 and -0.995 values between -0.99 and -0.905 values between -0.90 and -0.8110 values between -0.81 and -0.7314 values between -0.73 and -0.6422 values between -0.64 and -0.5622 values between -0.56 and -0.4738 values between -0.47 and -0.3953 values between -0.39 and -0.3082 values between -0.30 and -0.22156 values between -0.22 and -0.13439 values between -0.13 and -0.041313 values between -0.04 and 0.04860 values between 0.04 and 0.13497 values between 0.13 and 0.21220 values between 0.21 and 0.3076 values between 0.30 and 0.3827 values between 0.38 and 0.4714 values between 0.47 and 0.563 values between 0.56 and 0.643 values between 0.64 and 0.736 values between 0.73 and 0.814 values between 0.81 and 0.901 values between 0.90 and 0.981 values between 0.98 and 1.071 values between 1.07 and 1.161 values between 1.16 and 1.24offnegativepositiveoften

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. 还不如
  2. 究竟是
  3. 到底是
  4. falls
  5. jerne
  6. suất
  7. aurants
  8. гля
  9. fäh
  10. 究竟

Suppresses

  1. scientific
  2. scientists
  3. metab
  4. rep
  5. *\
  6. [mid
  7. movable
  8. #⏎⏎
  9. ubi
  10. implicitly