Neuron Explorer

Qwen3-0.6B

"until" after a negated clause

A neuron in layer 6 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 lights up on the word "until" in sentences like "it wasn't until 1775 that...", where something is said to have happened later than expected. Turning it up made the model's writing drift into repetitive or rambling text, which doesn't show a clear theme.

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

    980s, it wasn’t until the LIGO detector came online in 2002 that the hunt for

  2. #2 peak

    of Elizabeth I). However, it wasn’t until 1775 that Alexander Cummings added the u-bend and then

  3. #3 peak

    Up until the 1950s, the Folger also housed all nine of Gregory’s plaster reliefs in storage. It wasn’t until Dr. Frederick Hard—Scripps College president and Shakespeare scholar—negot

  4. #4 peak

    in 1531. In 1621, English colonists brought the first wild turkeys back to Massachusetts. However, it was not until after 1800 that domesticated turkeys became common in North America

  5. #5 peak

    as, 3 October 1997⏎While the flag of the Arapaho was first considered during the second World War, it was not until June 15, 1956 that one was adopted by the

  6. #6 peak

    founding of the University of Glasgow providing religious and educational status. It was not until the 16th century that Glasgow became prominent in world affairs. The city

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: Use of a black background, my own equipment, and the knowledge that I have used the computer Answer this question in the same row of rows Question: "Where is this?" Answer: "Here is a row of rows

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

    formality of the painting is indicative of the manifestation of abstraction and expressionism. Owning this liminal space of amalgamating tradition and contemporary, Akhlaq went on to influence future great artists like Shahzia Sikander, Rashid

  2. #2 peak

    of Pandit houses, asking them to convert, flee, or perish; as prominent Pandits were being murdered, the Chief Minister of the state Dr. Farooq Abdullah abandoned his people and ran away to London in order to let this slow

  3. #3 peak

    .org/10.1080/13527266.2011.620764⏎BELK, R. (201

  4. #4 peak

    This we have already explained in the introduction to Surah Al Muddaththir; and in E. N. 5 of Surah Al-Muzzammil also we have explained what great burden of the coming down of Revelation he had

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. L5:2957 suppresses
  2. L4:1675 excites
  3. L5:2390 suppresses
  4. L4:2882 suppresses
  5. L4:2393 weakly suppresses
  6. L5:781 weakly suppresses
  7. L4:661 weakly excites
  8. L5:2097 weakly excites
  9. L5:64 weakly excites
  10. L4:1632 weakly excites
  11. L4:1906 weakly suppresses
  12. L2:1259 weakly excites
  13. L5:1462 weakly excites
  14. L3:2705 weakly suppresses
  15. L4:1372 weakly suppresses
  16. L5:2393 weakly suppresses
  17. L5:2761 weakly excites
  18. L3:2620 weakly excites
  19. L5:681 weakly suppresses
  20. L2:42 weakly suppresses
  21. L4:88 weakly suppresses
  22. L4:2460 weakly suppresses 'City' in 'New York City'
  23. L5:1103 weakly suppresses
  24. L5:332 weakly suppresses

Feeds later layers

  1. L8:1869 suppresses
  2. L15:1386 excites prepositions after passive verbs
  3. L7:2970 suppresses
  4. L14:2150 excites
  5. L15:958 excites
  6. L9:2681 weakly excites
  7. L13:831 weakly excites
  8. L13:1515 weakly excites "your" before doctor or health provider
  9. L23:1345 weakly suppresses
  10. L7:1852 weakly excites the phrase 'by the time'
  11. L20:1612 weakly excites
  12. L23:2676 weakly suppresses
  13. L12:652 weakly excites
  14. L11:812 weakly excites
  15. L10:394 weakly excites
  16. L13:1162 weakly excites
  17. L11:1327 weakly excites
  18. L17:1995 weakly excites
  19. L17:2902 weakly suppresses
  20. L8:1962 weakly suppresses
  21. L7:2904 weakly excites
  22. L8:2420 weakly excites
  23. L7:507 weakly suppresses
  24. L19:883 weakly excites

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 -2.11 and -1.941 values between -1.94 and -1.781 values between -1.45 and -1.2814 values between -0.62 and -0.45126 values between -0.45 and -0.29667 values between -0.29 and -0.121836 values between -0.12 and 0.04974 values between 0.04 and 0.21227 values between 0.21 and 0.3732 values between 0.37 and 0.5411 values between 0.54 and 0.704 values between 0.70 and 0.874 values between 0.87 and 1.043 values between 1.04 and 1.203 values between 1.20 and 1.371 values between 1.53 and 1.701 values between 3.02 and 3.191 values between 4.01 and 4.181 values between 4.34 and 4.51offnegativepositiveoften

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. ILE
  2. leston
  3. enses
  4. fully
  5. ды
  6. ile
  7. ustin
  8. icias
  9. Reuters
  10. ество

Suppresses

  1. 协办
  2. aches
  3. uwagę
  4. dõi
  5. atted
  6. 伤心
  7. ouce
  8. Mahmoud
  9. dislikes
  10. 威胁