Neuron Explorer

Qwen3-0.6B

unclear: weak, varied mid-sentence words

A neuron in layer 14 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 responds only weakly and to a mix of words in the middle of explanatory sentences, such as "selecting the [feature]" or "motorists should [slow]". No single pattern stands out, and the steered text did not show a clear theme.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    data until a stopping criterion is met. The goal is to find the best split that maximizes the predictive accuracy of the model. This is done by selecting the feature that provides the most information gain or reduces the impurity of the data.⏎D

  2. #2 peak

    the Toolbar on the right (press T if you don’t see it) and under Rigid Body Tools, select ‘Bake To Keyframes’. Enter how many frames you want keys to<|endoftext|>Ann Wright: Deer in the garden⏎In the

  3. #3 peak

    00. The SDS preparers may also include additional information in various section(s).⏎Sections 1 through 8 contain general information about the chemical, identification, hazards, composition, safe handling practices, and emergency control measures (e.g., firefight

  4. #4 peak

    The high level of tetracycline resistance was maintained in spite of the declining concentration of tetracycline. (C) 2002 Elsevier Science B.V. All rights reserved.⏎|Publication status||Published - 2

  5. #5 peak

    Spreadsheets in Education (eJSiE):⏎2, Article 1.⏎Available at: http://epublications.bond.edu.au/ejsie/vol8/iss2/1<|endoftext|>Some may argue that sleeping is one of

  6. #6 peak

    sometimes called tipping moments—during the process-layout and die-design phases of a project. Unfortunately, many tooling engineers and die-process engineers do not fully understand the impact of off-center loading on part quality

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 people who are left out in the days of the 11/12/31, in the 12/31/31, 12/31/3, 31/31

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

    of contrasting brilliant color and shade. Some galleries insist that artists stand alongside their works during shows, juxtaposing the art and the artist.⏎Quiz: What is juxtaposition?⏎- The act of placing two things next

  2. #2 peak

    adjectives, adverbs, conjunctions, and prepositions. Each section:⏎- Defines the part of speech and its usage⏎- Gives examples to illustrate the definition⏎- Provides a range of interactive exercises to develop and consolidate understanding⏎- Offers

  3. #3 peak

    spring. This reduces the amount of water available to grow crops.⏎- In some areas, rainfall events are occurring less often, but are more extreme when they do occur. Extreme rainfall can wash away soils and damage crops. The ground also cannot soak

  4. #4 peak

    a, “Introduction to Data Mining with Case Studies”, 2nd Edition, PHI.⏎Students will be able to:⏎1. Explain the concepts in data mining and KDD, recognizing issues in Data Mining⏎2. Practice the preprocessing operations of

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. L12:1651 suppresses
  2. L11:1581 excites
  3. L13:3008 excites
  4. L12:1618 excites
  5. L13:1837 excites
  6. L13:2406 suppresses
  7. L12:1490 suppresses
  8. L11:979 suppresses
  9. L11:3034 suppresses
  10. L12:427 excites
  11. L12:822 excites
  12. L12:1701 excites
  13. L11:1136 suppresses
  14. L11:1151 excites
  15. L11:1887 excites
  16. L13:1394 suppresses
  17. L13:406 excites
  18. L11:2603 suppresses
  19. L13:2677 suppresses
  20. L11:2679 suppresses
  21. L11:573 weakly excites
  22. L13:2407 weakly suppresses
  23. L12:1044 weakly excites
  24. L13:1451 weakly excites

Feeds later layers

  1. L15:1425 strongly excites
  2. L17:2911 excites
  3. L16:2880 excites
  4. L16:357 suppresses
  5. L15:273 suppresses
  6. L16:2247 excites
  7. L16:3060 excites
  8. L16:1427 suppresses
  9. L15:1924 excites
  10. L15:143 suppresses
  11. L27:289 excites
  12. L16:1541 excites
  13. L15:3040 excites
  14. L15:509 excites
  15. L17:2529 excites
  16. L16:1364 excites
  17. L19:1092 suppresses
  18. L24:205 excites
  19. L15:1467 weakly excites
  20. L15:1339 weakly excites
  21. L15:412 weakly suppresses
  22. L21:739 weakly excites 'both X and Y' pairing words
  23. L15:2853 weakly excites
  24. L17:1491 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.

2 values between -0.86 and -0.773 values between -0.58 and -0.487 values between -0.48 and -0.3955 values between -0.39 and -0.29179 values between -0.29 and -0.20484 values between -0.20 and -0.101169 values between -0.10 and -0.011261 values between -0.01 and 0.09474 values between 0.09 and 0.19173 values between 0.19 and 0.2861 values between 0.28 and 0.3825 values between 0.38 and 0.477 values between 0.47 and 0.573 values between 0.57 and 0.662 values between 0.66 and 0.761 values between 0.76 and 0.851 values between 0.85 and 0.951 values between 2.85 and 2.95offnegativepositiveoften

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. 及其
  2. &&
  3. and
  4. &&
  5. selectAll
  6. 和支持
  7. ands
  8. agus
  9. それに
  10. 并向

Suppresses

  1. lyn
  2. iful
  3. 也可
  4. ="--
  5. -enter
  6. ulti
  7. レビ
  8. visor
  9. iese
  10. взгляд