Neuron Explorer

Qwen3-0.6B

"to" after "continues" or "continued"

A neuron in layer 1 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 "to" when it follows "continues" or "continued", as in "continues to develop". Turning it up produced repetitive or unrelated text, so the steering does not confirm it.

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

    attributes. Each node in the tree represents a decision based on a feature, and the tree continues to split the data until a stopping criterion is met. The goal is to find the

  2. #2 peak

    the young pastor (1 Timothy 1:18). God continues to use children today in their schools, homes, communities, and nations to testify of

  3. #3 peak

    footprint.⏎Advantages and disadvantages of using Sunpower solar panels⏎The use of solar energy is becoming increasingly popular across the globe as the technology continues to develop and prices drop. One of the most popular types of solar energy is Sun

  4. #4 peak

    ."<|endoftext|>When the ambient temperature descends below 32°F (0°C) rubber begins to lose its elasticity. As the temperature falls, the rubber continues to lose flexibility and its ability to resist abrasion, impact, and cutting. Eventually

  5. #5 peak

    for the concatenated component, although the data is still recoverable from errors because the parity is used during the component I/O. The resulting RAID 5 volume continues to handle a single component failure.⏎Concatenated components also differ in the sense that

  6. #6 peak

    edition has been released, and it continues to provide a key source of data, context and actionable insights for the U.S.

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 get the right to get the right to get the right to get the right to get the right to get the right to get the right to get the right to get the right to get the right to get 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

    MET protein.⏎MET is a receptor tyrosine kinase (RTK) that is produced as a single-chain precursor. The precursor is proteolyt

  2. #2 peak

    ize them so that the memory access runs, the R2 operation is held in an Instruction Pool pending completion of the first operation, and R5 and<|endoftext|>ATLANTA (AP) — A recent study shows one in four high school students drink

  3. #3 peak

    responsible for 17 percent of the protein we consume<|endoftext|>The first non-stop trans-Atlantic flight⏎THE VERY FIRST NON-STOP TRANS-ATLANTIC FLIGHT⏎Note: The U.S. Navy had already crossed the

  4. #4 peak

    the rainy season from May to December, it is advised that you rent or hire a four<|endoftext|>Palaeographical Importance of Nandinagari⏎BY: SATKARI MUKHOPADHYAYA⏎Jun

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. L0:1301 suppresses
  2. L0:178 suppresses
  3. L0:1224 suppresses
  4. L0:1 suppresses
  5. L0:873 weakly suppresses
  6. L0:2991 weakly suppresses
  7. L0:275 weakly suppresses
  8. L0:799 weakly suppresses
  9. L0:1858 weakly excites
  10. L0:902 weakly suppresses
  11. L0:499 weakly excites
  12. L0:1580 weakly excites
  13. L0:2854 weakly suppresses
  14. L0:2025 weakly suppresses
  15. L0:263 weakly suppresses
  16. L0:1672 weakly suppresses
  17. L0:819 weakly excites
  18. L0:157 weakly suppresses
  19. L0:196 weakly suppresses
  20. L0:1920 weakly excites
  21. L0:2623 weakly suppresses
  22. L0:2658 weakly suppresses
  23. L0:608 weakly suppresses
  24. L0:1153 weakly excites

Feeds later layers

  1. L8:3067 weakly excites
  2. L18:1137 weakly suppresses
  3. L3:981 weakly suppresses
  4. L17:1340 weakly excites
  5. L8:2475 weakly excites
  6. L8:1021 weakly suppresses
  7. L27:753 weakly excites
  8. L27:904 weakly excites
  9. L9:1876 weakly excites
  10. L12:3044 weakly excites
  11. L13:1080 weakly excites
  12. L15:1002 weakly suppresses periods ending claim sentences
  13. L9:2058 weakly excites
  14. L9:155 weakly excites
  15. L20:1013 weakly suppresses
  16. L20:2174 weakly excites
  17. L8:2710 weakly suppresses
  18. L4:399 weakly excites
  19. L6:639 weakly excites
  20. L18:2205 weakly excites
  21. L5:2567 weakly excites
  22. L6:900 weakly suppresses
  23. L11:2023 weakly suppresses
  24. L13:141 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.

36 values between -0.33 and -0.21255 values between -0.21 and -0.092670 values between -0.09 and 0.03859 values between 0.03 and 0.1570 values between 0.15 and 0.2712 values between 0.27 and 0.382 values between 0.38 and 0.501 values between 0.86 and 0.981 values between 1.22 and 1.341 values between 1.58 and 1.701 values between 4.33 and 4.45offnegativepositiveoften

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. **:
  2. :**
  3. 给她
  4. stands
  5. irs
  6. →put
  7. channelId
  8. associative
  9. ():⏎
  10. herself

Suppresses

  1. urger
  2. ecycle
  3. коп
  4. gie
  5. tics
  6. dock
  7. aic
  8. sleep
  9. .experimental
  10. ⏎ ⏎