Neuron Explorer

Qwen3-0.6B

'from' after 'differ' or 'vary'

A neuron in layer 17 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

The neuron lights up on the word 'from' when it follows words like 'differ', 'vary' or 'passed on', as in 'vary from person to person'. Turning it up made the model's writing loop on repeated words like 'country' and 'year', so the steering itself tells us little.

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

    incidence and can be a major concern for patients. The pain associated with dentin hypersensitivity is usually brought on by an external stimulus, and the intensity can differ from patient to patient. The diagnosis of dentin hypersensitivity often poses a challenge for

  2. #2 peak

    Stereotypes Are a Real Time-Saver across its essay about victorian period (see Fig. 5).⏎Definition Essay on SuccessThe definition of success differs from one person to another.⏎please use the sourses that I prepared for you.

  3. #3 peak

    Yule, Christmas, Sol Invictus, or Hogmanay, the winter season is typically a time of celebration around the world. Traditions vary widely from one country to the next,

  4. #4 peak

    halted, however, when James Kirk took command, and ordered a pursuit of the Narada toward Earth. (Star Trek)<|endoftext|>It's just fluctuating wildly from year to year as it always has done but by making some complex "allowances

  5. #5 peak

    edrals which slowly rose – sometimes taking hundreds of years to complete – in almost every land. Masons, wood-carvers, and other artists, wandered from country to country seeking employment, constructing a chapel here, painting a fresco there

  6. #6 peak

    allergy, the removal of the cause can be much more difficult.⏎There are some common triggers for hives. While it varies from individuals, these may include:⏎- Allergens (e.g., foods,

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 has been, in the last, and in the last, and in the last, I has, in the last, — in the last, — in the last, — in the last, — in the — — — — in

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

    .org/⏎is one of a series of problems designed to develop learners' team⏎working skills. Other tasks in the series can be found by going to⏎What are you aiming to⏎For the task:⏎The team has to

  2. #2 peak

    There⏎are two rows of seats in the waiting room. Using the rules, where⏎are they all sitting? Can you find all the possible ways?⏎What is the best way to shunt these carriages so that each train⏎can continue

  3. #3 peak

    = 0x80;⏎P0 = 0x98;⏎P0 = 0xC0;⏎P0 = 0xff;⏎} while(1); // Endless loop<|endoftext|>Wisconsin Well Water: Planning Web-Based

  4. #4 peak

    water quality and promoting growth of harmful, sometimes invasive, aquatic plants. Lawn care chemicals also move through the food chain, becoming more concentrated and harming wildlife. What’s more, using herbicides and pesticides to tackle weeds and insects can actually be counter

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. L12:2418 suppresses
  2. L11:2586 weakly suppresses
  3. L16:2979 weakly suppresses
  4. L15:791 weakly excites
  5. L8:296 weakly excites 'of' inside fixed noun phrases
  6. L5:94 weakly suppresses
  7. L13:1813 weakly suppresses
  8. L5:851 weakly suppresses
  9. L15:2318 weakly excites
  10. L16:2225 weakly suppresses
  11. L6:16 weakly excites
  12. L4:237 weakly excites
  13. L8:2650 weakly excites
  14. L8:2484 weakly suppresses
  15. L16:2326 weakly excites
  16. L0:1301 weakly suppresses
  17. L10:1594 weakly suppresses
  18. L7:805 weakly excites
  19. L6:2945 weakly suppresses
  20. L1:2575 weakly suppresses
  21. L13:965 weakly excites
  22. L4:350 weakly excites
  23. L0:546 weakly suppresses
  24. L15:292 weakly excites

Feeds later layers

  1. L19:2468 excites
  2. L18:3057 excites
  3. L19:2500 excites
  4. L20:1926 excites
  5. L19:1902 excites
  6. L18:1829 excites
  7. L22:2368 weakly excites
  8. L21:498 weakly excites
  9. L19:252 weakly excites
  10. L19:1042 weakly excites
  11. L19:1752 weakly excites
  12. L25:1583 weakly excites
  13. L18:2629 weakly excites
  14. L20:1497 weakly excites
  15. L23:465 weakly excites
  16. L25:224 weakly excites
  17. L22:529 weakly excites
  18. L27:580 weakly excites
  19. L18:158 weakly excites
  20. L20:1807 weakly excites
  21. L25:525 weakly excites
  22. L22:755 weakly excites 'all' or 'any' in 'people of all ages'
  23. L27:674 weakly suppresses
  24. L25:2129 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.

3 values between -1.79 and -1.445 values between -1.44 and -1.0821 values between -1.08 and -0.73356 values between -0.73 and -0.381884 values between -0.38 and -0.031411 values between -0.03 and 0.33136 values between 0.33 and 0.6829 values between 0.68 and 1.039 values between 1.03 and 1.3810 values between 1.38 and 1.747 values between 1.74 and 2.0914 values between 2.09 and 2.445 values between 2.44 and 2.791 values between 2.79 and 3.143 values between 3.14 and 3.501 values between 3.85 and 4.205 values between 4.20 and 4.551 values between 4.55 and 4.911 values between 4.91 and 5.261 values between 5.26 and 5.611 values between 5.96 and 6.321 values between 6.32 and 6.671 values between 10.19 and 10.541 values between 10.90 and 11.251 values between 11.95 and 12.31offnegativepositiveoften

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. sectors
  2. country
  3. species
  4. rier
  5. between
  6. boundaries
  7. atego
  8. year
  9. Between
  10. anoia

Suppresses

  1. 史
  2. getattr
  3. emin
  4. 芷
  5. .addAction
  6. assertTrue
  7. getAddress
  8. 重中之重
  9. حق
  10. getProperty