Neuron Explorer

Qwen3-0.6B

'says' after a quote

A neuron in layer 14 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 'says' when it comes right after a quotation and introduces who said it, as in news stories. Turning it up gave unrelated text, so the steering does not confirm the 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

    funded by the N.C. Commercial Fishing Resource Fund Grant Program.⏎“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.

  2. #2 peak

    tardigrades are vulnerable to high temperatures, though it seems that these critters would be able to acclimatize to increasing temperatures in their natural habitat,” says Ricardo Nieves, an author of<|endoftext|>January 10, 20

  3. #3 peak

    was first noted, a number of misconceptions, myths, and dogmas have arisen regarding fundamental aspects of boll weevil ecology in the subtropics,” says Showler. “These include its food sources, overwintering, and di

  4. #4 peak

    DCIS.⏎“The transition from DCIS to invasive ductal cancer is a key event in breast cancer progression, but it remains poorly understood,” says principal investigator Dr. Carlo M. Croce, director of Ohio State´s Human

  5. #5 peak

    to the holy grail of foot-and-mouth vaccines. Unlike the traditional vaccines, there is no chance that the empty shell vaccine could revert to an infectious form,' says Professor Stuart.⏎Dr Charleston adds: 'The ability to produce a vaccine outside of

  6. #6 peak

    Earth's surface. But when researchers looked at the western foothills southeast of Fresno, "we were no longer able to identify where the Moho was present," says Hersh Gilbert, a co-author

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 if this? well, what is the value of a different approach to the problem of the question, and what is called this? this. for what, and, and so, and, and, and, and, and, and

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

    fulness in children has been associated with obstinacy — a refusal to bend and comply with the wishes of a superior. We assume that our children

  2. #2 peak

    eltis polypodioides. It gets its name from its effective, defense mechanism against droughts. Being able to survive decades without water and still be able to return to a healthy state after some watering. This has made a lot of

  3. #3 peak

    are SDG 11 and SDG 16. SDG 11 purposes to “make cities inclusive, safe, resilient and sustainable” and SDG 16 seeks to “promote just, peaceful and inclusive societies”

  4. #4 peak

    firsthand experience of paying bribes or using a contact to get a job done in public office.⏎In 2012 India has ranked 94th out of 176 countries in Transparency International's Corruption Perceptions Index, tied

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. L7:7 suppresses
  2. L7:8 suppresses
  3. L13:3 suppresses
  4. L6:325 suppresses
  5. L6:60 weakly suppresses
  6. L7:135 weakly suppresses 'of' after 'issue' in citations
  7. L10:43 weakly excites "terms" in "in terms of"
  8. L13:1000 weakly suppresses
  9. L6:443 weakly excites
  10. L6:823 weakly suppresses
  11. L9:143 weakly suppresses
  12. L8:319 weakly excites
  13. L9:35 weakly suppresses
  14. L9:330 weakly excites gap where a number was removed
  15. L13:338 weakly suppresses
  16. L13:866 weakly excites
  17. L8:187 weakly suppresses
  18. L9:933 weakly suppresses
  19. L10:458 weakly excites
  20. L7:1723 weakly excites
  21. L1:2604 weakly suppresses
  22. L13:1372 weakly suppresses number before an age or week count
  23. L6:562 weakly suppresses
  24. L5:248 weakly excites

Feeds later layers

  1. L18:1014 excites
  2. L15:1601 weakly excites
  3. L15:248 weakly excites
  4. L20:48 weakly excites
  5. L16:2600 weakly excites
  6. L16:326 weakly excites
  7. L19:152 weakly suppresses
  8. L16:2521 weakly excites
  9. L18:2448 weakly suppresses
  10. L16:1135 weakly excites
  11. L16:292 weakly excites
  12. L23:2043 weakly suppresses
  13. L24:511 weakly excites
  14. L23:2522 weakly suppresses
  15. L16:662 weakly excites
  16. L15:576 weakly suppresses
  17. L24:1395 weakly excites
  18. L17:2057 weakly excites
  19. L22:832 weakly suppresses
  20. L15:1524 weakly excites
  21. L18:1837 weakly suppresses
  22. L16:2182 weakly excites
  23. L18:2529 weakly excites
  24. L23:1922 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.

2 values between -5.06 and -4.771 values between -4.77 and -4.473 values between -4.18 and -3.888 values between -3.58 and -3.295 values between -3.29 and -2.9912 values between -2.99 and -2.707 values between -2.70 and -2.4019 values between -2.40 and -2.1027 values between -2.10 and -1.8128 values between -1.81 and -1.5165 values between -1.51 and -1.2293 values between -1.22 and -0.92157 values between -0.92 and -0.63313 values between -0.63 and -0.33621 values between -0.33 and -0.031059 values between -0.03 and 0.26719 values between 0.26 and 0.56394 values between 0.56 and 0.85197 values between 0.85 and 1.1590 values between 1.15 and 1.4548 values between 1.45 and 1.7422 values between 1.74 and 2.048 values between 2.04 and 2.336 values between 2.33 and 2.631 values between 2.63 and 2.921 values between 2.92 and 3.221 values between 3.22 and 3.521 values between 6.47 and 6.77offnegativepositiveoften

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. onis
  2. ingly
  3. ronics
  4. elon
  5. ombok
  6. 不代表
  7. URL
  8. 还有很多
  9. uess
  10. oring

Suppresses

  1. 剽
  2. sinking
  3. ожно
  4. riot
  5. WHEN
  6. 每
  7. BAM
  8. utow
  9. тра
  10. 体内