Neuron Explorer

Qwen3-0.6B

open parenthesis after a name or term

A neuron in layer 0 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 reacts to an opening bracket that comes right after a name, term or date, as in 'cognitive behavioural therapy (CBT)'. It is the bracket that introduces an abbreviation or a date. Turning it up did not clearly change what the model wrote.

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

    therapy (CBT), focuses on changing negative content of thoughts while MBCT emphasizes the process

  2. #2 peak

    Karnak (this name deriving from the Arab village of Al-karnak). The ruins of

  3. #3 peak

    oleonic Wars (1803-1815) the colony passed into the control

  4. #4 peak

    PHIA (CBS) – So, that’s why they say to pinch back plants by the

  5. #5 peak

    lin Road Gaol (1843-45) and the County Courthouse (18

  6. #6 peak

    lifestyle (meeting 0-2 lifestyle recommendations), those who were exposed to gestational diabetes

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 been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately

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

    small as asteroids go. Because of that (and its current distance of 450 million kilometers) it's a bit

  2. #2 peak

    ). It is otherwise usually rather plain, with dull olive upperparts, a dull grayish cap, and whitish underparts with (often) yellow on the flanks or undertail coverts. Each year,

  3. #3 peak

    a signature different from those of the recipient species, the analysis of local variations of signature along recipient genome may allow for detecting exogenous DNA. The strategy consists in (i) scanning the genome with a sliding window, and calculating the corresponding local signature

  4. #4 peak

    to, and recovery from drugs and alcohol.⏎We analyzed anonymous surveys completed by students before and after attending a DST performance, and transcripts of focus group interviews conducted with (1) program developers, (2) stakeholders, (3) performers, and

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

None: this is the first layer.

Feeds later layers

  1. L13:338 suppresses
  2. L17:652 weakly excites
  3. L9:2543 weakly suppresses
  4. L6:1914 weakly excites
  5. L9:737 weakly suppresses
  6. L14:88 weakly suppresses
  7. L16:714 weakly excites
  8. L1:604 weakly suppresses
  9. L14:609 weakly excites
  10. L23:2982 weakly excites
  11. L5:2297 weakly suppresses
  12. L2:108 weakly excites
  13. L8:2824 weakly suppresses
  14. L23:2967 weakly suppresses
  15. L13:1792 weakly suppresses
  16. L9:656 weakly suppresses
  17. L11:297 weakly excites
  18. L12:2924 weakly suppresses
  19. L2:263 weakly suppresses
  20. L8:2865 weakly excites
  21. L21:1177 weakly suppresses
  22. L11:101 weakly excites
  23. L9:1729 weakly suppresses
  24. L12:410 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 -1.01 and -0.961 values between -0.79 and -0.734 values between -0.73 and -0.684 values between -0.68 and -0.621 values between -0.62 and -0.573 values between -0.51 and -0.454 values between -0.45 and -0.403 values between -0.40 and -0.349 values between -0.34 and -0.2916 values between -0.29 and -0.2334 values between -0.23 and -0.18112 values between -0.18 and -0.12266 values between -0.12 and -0.06868 values between -0.06 and -0.011496 values between -0.01 and 0.05621 values between 0.05 and 0.10265 values between 0.10 and 0.16102 values between 0.16 and 0.2159 values between 0.21 and 0.2710 values between 0.27 and 0.3313 values between 0.33 and 0.384 values between 0.38 and 0.441 values between 0.44 and 0.491 values between 0.66 and 0.721 values between 0.77 and 0.833 values between 0.94 and 0.992 values between 1.05 and 1.112 values between 1.11 and 1.162 values between 1.16 and 1.22offnegativepositiveoften

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. tesy
  2. ees
  3. liest
  4. 一听
  5. uchi
  6. airs
  7. oley
  8. 一经
  9. pond
  10. auer

Suppresses

  1. jan
  2. {}",
  3. {j
  4. inho
  5. 肢
  6. rado
  7. eni
  8. ежду
  9. .Gray
  10. .increment