Neuron Explorer

Qwen3-0.6B

article 'a' after a role or title

A neuron in layer 23 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 "a" or "an" that comes right after a phrase like "and", "of which he was", or "not", where a job title or role is about to follow. Turning it up gave text that did not clearly relate to this pattern.

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

    at IBM Research – India for ten years. His work has led to twelve patents from the USPTO. A Senior Member of the IEEE and the ACM and a Fellow of the UK Higher Education Authority, he is the recipient of the INAE

  2. #2 peak

    “The goal of the research was to better understand the economic impact of North Carolina’s seafood industry,” says Jane Harrison, North Carolina Sea Grant’s coastal economics specialist and a project lead. “The shellfish results indicate the strength of the growing aquac

  3. #3 peak

    —still half the rate of the global average.1⏎Gayle Tzemach Lemmon, a senior fellow at the Council on Foreign Relations as well as a journalist and the best-selling author of The Dressmaker of Khair Khana,

  4. #4 peak

    Festivals, and Associate Producer/Writer for Westinghouse TV. He currently serves on the Executive Board of the Stage Managers’ Association, of which he was a founding member and former chairperson. He has taught at both Emerson College and Stock

  5. #5 peak

    has been sworn in as the new Director of the Central Intelligence Agency. As a career diplomat, he is the first person to serve in this role who is not a career intelligence officer. We believe that his experience and expertise will be a valuable asset

  6. #6 peak

    .<|endoftext|>Dr Ana Cruz is a Research Fellow with Modernising Medical Microbiology, and a key part of the new CRyPTIC Project, where we will aim to

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: what is the value of the least least least least least least least least least least least least least least least least least least least least least least least least least least least least least least least least least least least least least least least least least least

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

    activities of the Mennonites of Russia can be divided into three periods: 1789-1860, 1860-191

  2. #2 peak

    is a cousin of the former first lady, is the chief rabbi of the black Jewish community. He also leads Beth Shalom B’nai Zaken Ethiopian Hebrew Congregation in Chicago. Uniquely among Hebrew Israelite rabbis,

  3. #3 peak

    from which the priests wash their hands and feet before engaging in Divine service; the compounding of the incense and of the anointing oil; and the Shabbat. I shall focus here upon the washing of the hands.⏎Hand-w

  4. #4 peak

    to purchase a Torah scroll and ram’s horn for the upcoming High Holy Day services. Originally called the Little Rock Congregation, the group formally incorporated as B’nai Israel in 1867 and elected Morris Navra as its first president

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. L20:702 excites
  2. L20:2148 excites
  3. L19:1200 weakly excites
  4. L1:517 weakly suppresses
  5. L20:1020 weakly suppresses
  6. L4:2935 weakly suppresses
  7. L21:1148 weakly suppresses
  8. L11:2438 weakly suppresses 'from' after origin verbs like imported or brought
  9. L6:2871 weakly suppresses
  10. L3:3042 weakly suppresses
  11. L18:2451 weakly suppresses
  12. L22:1639 weakly suppresses
  13. L3:349 weakly excites
  14. L21:2041 weakly suppresses
  15. L19:1127 weakly excites
  16. L12:667 weakly suppresses
  17. L1:1826 weakly excites
  18. L21:1697 weakly suppresses
  19. L17:1343 weakly excites
  20. L15:666 weakly excites
  21. L9:747 weakly excites
  22. L5:2768 weakly excites 'to' after 'happen' (happen to)
  23. L20:2659 weakly suppresses
  24. L22:1341 weakly suppresses

Feeds later layers

  1. L27:1920 weakly excites
  2. L27:1125 weakly suppresses
  3. L27:842 weakly excites
  4. L25:1448 weakly suppresses
  5. L27:2227 weakly suppresses
  6. L27:2544 weakly excites
  7. L24:2913 weakly excites
  8. L24:485 weakly suppresses
  9. L24:1353 weakly suppresses
  10. L26:568 weakly suppresses
  11. L27:1719 weakly excites
  12. L25:1481 weakly excites
  13. L24:316 weakly excites
  14. L24:1631 weakly suppresses
  15. L25:497 weakly suppresses
  16. L27:1573 weakly excites
  17. L24:1441 weakly excites
  18. L26:1874 weakly excites
  19. L25:1825 weakly excites
  20. L25:2793 weakly suppresses
  21. L24:130 weakly excites line break at the end of a line
  22. L26:1936 weakly excites
  23. L25:1802 weakly excites
  24. L25:1361 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 -11.25 and -10.421 values between -10.42 and -9.591 values between -8.76 and -7.931 values between -7.10 and -6.276 values between -6.27 and -5.451 values between -5.45 and -4.628 values between -4.62 and -3.7912 values between -3.79 and -2.9628 values between -2.96 and -2.1360 values between -2.13 and -1.30266 values between -1.30 and -0.472644 values between -0.47 and 0.36563 values between 0.36 and 1.19129 values between 1.19 and 2.0248 values between 2.02 and 2.8545 values between 2.85 and 3.6823 values between 3.68 and 4.5117 values between 4.51 and 5.3414 values between 5.34 and 6.178 values between 6.17 and 7.008 values between 7.00 and 7.8210 values between 7.82 and 8.654 values between 8.65 and 9.482 values between 9.48 and 10.312 values between 10.31 and 11.141 values between 11.14 and 11.971 values between 11.97 and 12.801 values between 12.80 and 13.631 values between 13.63 and 14.461 values between 14.46 and 15.291 values between 21.09 and 21.92offnegativepositiveoften

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. essian
  2. ci
  3. 自然
  4. educt
  5. 外籍
  6. (IConfiguration
  7. �
  8. 冮
  9. 과정
  10. natural

Suppresses

  1. istol
  2. engers
  3. letic
  4. mania
  5. ista
  6. eah
  7. arious
  8. .setAuto
  9. uales
  10. obbies