Neuron Explorer

Qwen3-0.6B

firsthand accounts and testimony

A neuron in layer 5 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 lights up on words about first-hand accounts, such as "testimonies", "first-person", "firsthand" and "eyewitness". When it was turned up, the model's text did not clearly follow this theme, so the steering results don't add much.

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

    War narratives exist and of these, Blight notes, "a mere handful" are first-person accounts by slaves who ran away and freed themselves.Skip to next paragraph⏎Subscribe Today

  2. #2 peak

    slavery is injustice. Frederick Douglass, a black slave previously, witnessed the Inhumane slavery tragedies and therefore able to unfurled his narrative In a first-person perspective and a chronological timeline. More importantly, he was one of the very few

  3. #3 peak

    acht from hundreds of Jews who had fled, but Hartshorne joined the Secret Service shortly afterwards and the accounts he gathered were forgotten - until now. These eyewitness testimonies - published here for the first time, with a foreword by Saul Fried

  4. #4 peak

    interested in modern history, Nazi Germany and the historical experience of the Jews.⏎About Uta GerhardtSee more books from this Author⏎The value of these testimonies lies above all in their detail and immediacy.Read Full Review of The Night

  5. #5 peak

    of the most extensive in terms of the number of publications. All kinds of stories and accounts are told here, from the autobiographical and biographical, the testimonies of individuals and groups of people, to surveys of countries and continents through the lenses

  6. #6 peak

    break. The result is an approach-avoidance conflict whereby the habit becomes more compulsive and the gambler more defensive and increasingly confused and desperate. The personal testimonies of addicted gamblers and their family members have long been witness to the power

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: the idea of a single pizza. It is a beautiful, simple concept. But no matter what, a pizza. Okay, what is the solution? This is the pizza. Just a pizza. Alright, I need to solve this problem.

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

    on one page (FILE, PAGE SETUP or FILE, PRINTER SETUP in most browsers).⏎Printable version of these instructions<|endoftext|>By now, you’ve probably heard of the replication crisis in science. In summary, many conclusions from experiments done in a

  2. #2 peak

    God?⏎2. Find out what the apocryphal Gospels are and of what value they are.⏎3. Why did not Mary after the Annunciation run to the Jewish priests and tell all about the visit of the angel, and

  3. #3 peak

    recipes or instructions. In many cases, content like photos or videos is shared. Read on to hear more about the more popular options.⏎You have probably already heard of Facebook. It has come a long way since its early days as a directory for

  4. #4 peak

    on www.tab<|endoftext|>Sun Protection 101: SPF, UPF, Sun-Protective Clothing & Sunscreen Explained⏎By now you’ve heard of shirts, hats, and various other items of apparel all designed with UPF treatment

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. L3:2324 excites
  2. L3:758 weakly suppresses
  3. L3:2667 weakly suppresses
  4. L3:2909 weakly excites
  5. L2:3042 weakly excites
  6. L4:1577 weakly suppresses
  7. L4:23 weakly excites
  8. L3:2275 weakly suppresses
  9. L4:3052 weakly excites
  10. L1:666 weakly suppresses
  11. L0:56 weakly excites
  12. L0:312 weakly excites
  13. L4:2921 weakly excites
  14. L4:1600 weakly suppresses
  15. L0:6 weakly suppresses
  16. L2:75 weakly suppresses
  17. L3:1426 weakly excites
  18. L3:626 weakly excites
  19. L1:1674 weakly excites
  20. L1:1575 weakly suppresses
  21. L3:477 weakly suppresses
  22. L1:2880 weakly suppresses
  23. L3:1737 weakly excites
  24. L2:2376 weakly excites

Feeds later layers

  1. L8:1508 excites
  2. L9:2877 excites
  3. L7:479 excites
  4. L9:1871 weakly excites
  5. L16:1054 weakly excites
  6. L6:1547 weakly suppresses
  7. L7:2429 weakly excites
  8. L14:275 weakly suppresses
  9. L6:1559 weakly suppresses
  10. L18:2865 weakly excites
  11. L8:2913 weakly excites
  12. L11:2605 weakly excites
  13. L9:1747 weakly suppresses
  14. L12:597 weakly excites
  15. L16:1341 weakly suppresses
  16. L8:2692 weakly suppresses
  17. L11:2043 weakly excites
  18. L16:2627 weakly excites
  19. L22:1036 weakly excites
  20. L7:3009 weakly suppresses
  21. L10:2355 weakly suppresses
  22. L22:1838 weakly suppresses
  23. L22:2053 weakly excites
  24. L22:1357 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 -1.69 and -1.581 values between -1.58 and -1.482 values between -0.75 and -0.644 values between -0.64 and -0.549 values between -0.54 and -0.4334 values between -0.43 and -0.33125 values between -0.33 and -0.22454 values between -0.22 and -0.121161 values between -0.12 and -0.011378 values between -0.01 and 0.10519 values between 0.10 and 0.20144 values between 0.20 and 0.3139 values between 0.31 and 0.4114 values between 0.41 and 0.527 values between 0.52 and 0.622 values between 0.62 and 0.732 values between 0.73 and 0.831 values between 0.83 and 0.942 values between 0.94 and 1.042 values between 1.04 and 1.153 values between 1.25 and 1.361 values between 1.36 and 1.461 values between 1.67 and 1.781 values between 1.99 and 2.091 values between 2.41 and 2.51offnegativepositiveoften

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. testimon
  2. firsthand
  3. testimony
  4. confirmed
  5. ыта
  6. orque
  7. нят
  8. footage
  9. 亲身
  10. ount

Suppresses

  1. agram
  2. 放过
  3. complexes
  4. wrappers
  5. selects
  6. igma
  7. glGen
  8. agrams
  9. rama
  10. lig