Neuron Explorer

Qwen3-0.6B

unclear: scattered mid-phrase word pieces

A neuron in layer 11 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on a grab bag of word pieces inside longer words or phrases, such as the end of 'Simpson' or the middle of 'leach', with no clear shared topic. Turning it up gave rambling, unrelated text, so it is hard to say what it does.

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

    of soda, vaseline, or burnt with a match in an attempt to remove. Such action can induce a tick to release spirochetes and/or other pathogens known to cause

  2. #2 peak

    to hurricane.⏎ONCE WE CALL IT A HURRICANE, HOW DO WE CATEGORIZE IT?⏎We look to the Saffir-Simpson Hurricane scale, developed as a classification system for Western Hemisphere tropical cyclones in the

  3. #3 peak

    processes related to WTO Agreements such as agriculture, sanitary and phytosanitary measures (SPS), technical barriers to trade (TBT), trade-related aspects of intellectual property rights

  4. #4 peak

    When should I call my healthcare provider?⏎If your symptoms have not improved within the time frame suggested by your healthcare provider, you should let them know. Also, if your symptoms get worse or you get new symptoms,

  5. #5 peak

    age), you can feed them strawberries. However, the strawberries should be chopped or sliced to avoid choking. They also need rinsing to wash off pesticides or chemical residues.⏎Can goats eat strawberry tops?⏎Yes, they can. Strawberry tops are not

  6. #6 peak

    in different types of plastic. The monomers used to make PET bottles are derived from petroleum, and there have been some concerns that chemicals from the plastic could leach into the contents of the bottle, potentially contaminating the liquid inside.⏎However,

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 been thinking about the fact that the number of the number of the number of the number of the number of the number of the number of the number of the number of the number of the number of the number of the number of 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

    explanations. Some radar operators and pilots insisted that the movements of the objects were too coordinated to be random atmospheric anomalies. Furthermore, skeptics pointed out that temperature inversions were common, but such radar sightings were not.⏎The 1952

  2. #2 peak

    AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file

  3. #3 peak

    disease is a serious condition that occurs when the arteries responsible for carrying blood to the heard become narrow due to the buildup of plaque and cholesterol inside the arteries. An angioplasty

  4. #4 peak

    D.C., at age 67.⏎In 1930, the chief justice of the United States, William Howard Taft, resigned for health reasons. (He died just over a month later.)⏎In 1943

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. L10:983 suppresses
  2. L10:1569 suppresses
  3. L10:1040 suppresses
  4. L10:1269 excites
  5. L7:2759 suppresses
  6. L6:2171 weakly suppresses
  7. L7:2111 weakly excites closures and construction
  8. L10:2444 weakly excites
  9. L2:3052 weakly suppresses
  10. L7:2523 weakly excites 'since' after a change or growth
  11. L3:2952 weakly excites
  12. L9:233 weakly excites
  13. L3:1531 weakly suppresses
  14. L10:1296 weakly excites
  15. L8:1909 weakly excites
  16. L7:1093 weakly excites
  17. L9:803 weakly excites
  18. L4:1225 weakly suppresses
  19. L9:1613 weakly excites
  20. L9:311 weakly suppresses
  21. L8:951 weakly suppresses
  22. L0:643 weakly suppresses
  23. L7:1534 weakly suppresses
  24. L7:1841 weakly excites

Feeds later layers

  1. L13:2804 excites
  2. L14:1459 suppresses
  3. L12:561 suppresses
  4. L13:174 suppresses
  5. L12:0 suppresses
  6. L21:2987 weakly excites
  7. L13:2368 weakly excites
  8. L12:2993 weakly excites
  9. L17:2777 weakly excites
  10. L13:3051 weakly excites
  11. L14:867 weakly excites
  12. L13:1215 weakly excites
  13. L13:560 weakly excites
  14. L12:20 weakly suppresses
  15. L13:596 weakly excites
  16. L16:1789 weakly excites
  17. L13:2618 weakly excites
  18. L17:828 weakly excites
  19. L22:1199 weakly excites
  20. L13:200 weakly suppresses
  21. L12:664 weakly suppresses
  22. L17:2972 weakly excites
  23. L12:281 weakly suppresses
  24. L16:636 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 -4.68 and -4.502 values between -3.61 and -3.431 values between -3.25 and -3.083 values between -3.08 and -2.901 values between -2.90 and -2.721 values between -2.72 and -2.541 values between -2.54 and -2.377 values between -2.37 and -2.198 values between -2.19 and -2.017 values between -2.01 and -1.8314 values between -1.83 and -1.6516 values between -1.65 and -1.4824 values between -1.48 and -1.3023 values between -1.30 and -1.1242 values between -1.12 and -0.9466 values between -0.94 and -0.77105 values between -0.77 and -0.59144 values between -0.59 and -0.41220 values between -0.41 and -0.23425 values between -0.23 and -0.051057 values between -0.05 and 0.12727 values between 0.12 and 0.30424 values between 0.30 and 0.48266 values between 0.48 and 0.66169 values between 0.66 and 0.8470 values between 0.84 and 1.0141 values between 1.01 and 1.1914 values between 1.19 and 1.3712 values between 1.37 and 1.557 values between 1.55 and 1.725 values between 1.72 and 1.903 values between 1.90 and 2.081 values between 2.08 and 2.261 values between 2.26 and 2.44offnegativepositiveoften

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. AppState
  2. mó
  3. ối
  4. Pasadena
  5. (Component
  6. フォー
  7. 僭
  8. 鳌
  9. 最先进的
  10. MASK

Suppresses

  1. bn
  2. otype
  3. بن
  4. hte
  5. oir
  6. 一二
  7. -z
  8. isc
  9. [--
  10. [Unit