Neuron Explorer

Qwen3-0.6B

table column header boundaries

A neuron in layer 8 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 double bar that separates a column heading from the next heading in pasted tables, such as the gap between 'Format' and 'File Header'. Turning it up made the model's output drift into unrelated text, with no 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

    any parameters. Other file formats can be read by creating an NMRView parameter file that contains a description of the layout of the file.⏎|Program||Format||File Header||Block Header||Parameters|⏎|Varian||Serial||32

  2. #2 peak

    Gulf, down 2.5 percent⏎|Region||December 2013||December 2012||Percent Change|⏎|Region||December 2013||November 2013||Percent Change

  3. #3 peak

    reports, he continued "wielding his machete until the foe had<|endoftext|>|Theme||Visible||Selectable||Appearance||Zoom Range (now: 0)|⏎Well

  4. #4 peak

    of Sukkot. But everywhere else, the following chart applies.⏎Fall 2001 Jewish Holiday Calendar⏎|Biblical Name||Modern Holiday||Hebrew Date||English Date|⏎|Feast of Trumpets (Lev

  5. #5 peak

    GNU Oflox⏎This is a GNU package:octave⏎released on 31 May 2012⏎|License||Verified by||Verified on||Notes|⏎|GPLv2||Kelly Hopkins||6 December

  6. #6 peak

    ARAB 300- Advanced Arabic⏎Arabic Long Term Schedule (updated Spring 2012)⏎|FALL 2012||ARAB 100|⏎|SPRING 2013||

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 a 4th year in college, and I want to know how to calculate the probability of getting a higher number of 3 in my 4th year. Let's see if I can calculate the probability of getting a higher

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

    , we could:⏎|Vitamin B12 Need||94%||Confirm|⏎|Lupus (SLE)||2%||Ruled out|⏎|A

  2. #2 peak

    Place of Publication||Perth, Western Australia|⏎|Number of pages||173|⏎|ISBN (Print)||9 7

  3. #3 peak

    FF’s Daniel Raimi gives a balanced and accessible view of oil and gas development, drawing on both scholarly literature and his t...<|endoftext|>|Full name||Noam Chomsky

  4. #4 peak

    17.0.0/17⏎- We need a network with <|endoftext|>|Classification||Electronic musical instrument, synthesizer|⏎|Inventor(s)||Suzuki Musical Instrument Corporation|⏎|Tronichord, autoh

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. L6:1056 suppresses
  2. L7:335 excites
  3. L5:1512 suppresses
  4. L5:894 suppresses
  5. L2:974 excites
  6. L7:63 weakly excites
  7. L6:1691 weakly suppresses
  8. L6:16 weakly excites
  9. L4:556 weakly excites
  10. L7:1717 weakly suppresses
  11. L7:69 weakly excites
  12. L6:234 weakly suppresses
  13. L5:50 weakly excites
  14. L7:1242 weakly suppresses
  15. L1:2801 weakly excites
  16. L2:916 weakly excites
  17. L6:2156 weakly excites
  18. L5:75 weakly excites
  19. L3:115 weakly suppresses
  20. L7:271 weakly suppresses
  21. L3:1962 weakly excites
  22. L0:2821 weakly excites
  23. L6:7 weakly suppresses
  24. L5:64 weakly excites

Feeds later layers

  1. L9:1066 excites
  2. L13:1850 excites
  3. L10:2609 excites
  4. L14:2303 weakly excites
  5. L9:1734 weakly excites
  6. L18:1900 weakly suppresses
  7. L11:2508 weakly excites
  8. L13:201 weakly excites
  9. L23:2766 weakly suppresses
  10. L14:3039 weakly excites
  11. L12:660 weakly excites
  12. L11:2507 weakly excites
  13. L10:1849 weakly excites
  14. L15:234 weakly excites
  15. L10:2219 weakly suppresses
  16. L20:2520 weakly suppresses
  17. L9:1638 weakly excites
  18. L16:1647 weakly suppresses
  19. L22:853 weakly suppresses
  20. L15:457 weakly excites
  21. L10:2862 weakly excites
  22. L23:1323 weakly excites
  23. L19:2291 weakly excites
  24. L18:3015 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 -0.78 and -0.731 values between -0.68 and -0.641 values between -0.64 and -0.591 values between -0.59 and -0.545 values between -0.54 and -0.5015 values between -0.50 and -0.4515 values between -0.45 and -0.4033 values between -0.40 and -0.3541 values between -0.35 and -0.3151 values between -0.31 and -0.2684 values between -0.26 and -0.21146 values between -0.21 and -0.17265 values between -0.17 and -0.12416 values between -0.12 and -0.07672 values between -0.07 and -0.031053 values between -0.03 and 0.02601 values between 0.02 and 0.07261 values between 0.07 and 0.12117 values between 0.12 and 0.1657 values between 0.16 and 0.2132 values between 0.21 and 0.2620 values between 0.26 and 0.307 values between 0.30 and 0.358 values between 0.35 and 0.401 values between 0.40 and 0.441 values between 0.59 and 0.631 values between 0.77 and 0.822 values between 1.06 and 1.10offnegativepositiveoften

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. amba
  2. 门前
  3. ��态
  4. ,key
  5. abbo
  6. atable
  7. acted
  8. ='.
  9. rama
  10. =key

Suppresses

  1. illery
  2. erville
  3. ọi
  4. eros
  5. глав
  6. qrt
  7. 伍
  8. 辜负
  9. UTURE
  10. éré