Neuron Explorer

Qwen3-0.6B

the word 'balance' in many senses

A neuron in layer 25 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 lights up on the word 'balance', as in balance sheet, balance of power, or a proper balance between things. When it was turned up, the text drifted into odd, unrelated word salad, so its role is not clear from that test.

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

    into the reporting side of accounting, and in order to keep this article as simple as possible, we are not going to expand on the relationship of deferred items to balance sheets, profit and loss sheets

  2. #2 peak

    .⏎Why it Matters:⏎Any entity which needs to generate and use capital, whether it be a bank, a household, or a business, must have the proper balance between income producing assets and liquid assets such as cash which produce no income. Holding

  3. #3 peak

    and Production Projections⏎- Outlook for Natural Gas Prices⏎- Economic Benefits and Implications⏎- Public Health and the Environment⏎States Take Action: The Balancing Act⏎State Policy Actions⏎- Increasing Transparency⏎- Generating Revenue Through Severance

  4. #4 peak

    Understanding Bilan Comptable: A Comprehensive Guide⏎Understanding Bilan Comptable: A Comprehensive Guide⏎A bilan comptable, also known as a balance sheet, is a fundamental financial statement that provides a snapshot of a company

  5. #5 peak

    day it was struck. - (LC.025)<|endoftext|>The fur trade caused Native people to become partially dependent on European trade goods, changing their traditional balance between needs and resources. The fur traders introduced alcohol, causing social and health problems

  6. #6 peak

    Preparing Financial Statements⏎• Understanding Double Entry Concepts⏎• Recording Transactions- Journals, Cash Book, General Ledger⏎• Understanding the differences of Income Statement & Balance Sheet and their roles⏎Assumptions –Sources of information⏎• Understanding Relevant

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: New-becun-becun-becun-becun-becun-becun-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b

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

    borrower's debt-to-income ratio. To prevent it from affecting the stress test, the borrower can make efforts to pay off their existing loans or reduce their outstanding debt before applying for a mortgage loan.⏎If the borrower already holds property, the

  2. #2 peak

    a savings account.⏎Once the “long-term savings” envelope reaches a certain dollar amount (you can agree upon this amount with your child beforehand), it’s time

  3. #3 peak

    0% on ads, and your contributions will help us continue providing the best free cultural and educational materials to learners everywhere. You can contribute through PayPal, Patreon, Venmo (@openculture) and Crypto. Thanks!⏎Gertrude Stein Gets

  4. #4 peak

    bond price formula⏎Bond price is calculated as the present value of the cash flow generated by the bond, namely the coupon payment throughout the life of the bond and the principal payment, or the balloon payment, at the end of the bond's life

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. L23:3069 excites
  2. L22:2706 suppresses
  3. L18:2554 excites
  4. L22:408 suppresses
  5. L22:1267 excites
  6. L3:1506 suppresses
  7. L9:2632 excites
  8. L4:2237 excites
  9. L6:2375 suppresses
  10. L16:3010 excites
  11. L6:1145 excites
  12. L23:2342 weakly suppresses
  13. L20:630 weakly suppresses
  14. L21:2915 weakly suppresses
  15. L21:1563 weakly suppresses
  16. L11:2768 weakly excites
  17. L3:1094 weakly excites
  18. L0:587 weakly excites
  19. L3:1446 weakly excites
  20. L4:2606 weakly suppresses
  21. L9:2808 weakly suppresses
  22. L2:2439 weakly excites
  23. L23:1630 weakly excites
  24. L24:2372 weakly excites

Feeds later layers

  1. L27:2899 strongly suppresses
  2. L27:2860 weakly suppresses
  3. L26:1458 weakly suppresses
  4. L26:1099 weakly suppresses
  5. L27:2160 weakly suppresses
  6. L26:2502 weakly suppresses
  7. L26:365 weakly excites
  8. L27:2530 weakly suppresses
  9. L27:189 weakly excites
  10. L27:2045 weakly excites
  11. L26:2815 weakly suppresses
  12. L27:699 weakly suppresses
  13. L26:2017 weakly excites
  14. L27:368 weakly suppresses
  15. L27:2130 weakly suppresses
  16. L27:2746 weakly suppresses
  17. L27:1660 weakly suppresses
  18. L27:2974 weakly suppresses
  19. L27:2911 weakly suppresses
  20. L27:1109 weakly excites
  21. L26:2758 weakly suppresses
  22. L26:3055 weakly suppresses
  23. L27:614 weakly suppresses
  24. L27:618 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 -19.84 and -19.001 values between -19.00 and -18.161 values between -15.64 and -14.801 values between -13.11 and -12.272 values between -11.43 and -10.592 values between -10.59 and -9.751 values between -8.07 and -7.231 values between -7.23 and -6.383 values between -6.38 and -5.544 values between -5.54 and -4.703 values between -4.70 and -3.867 values between -3.86 and -3.0213 values between -3.02 and -2.1823 values between -2.18 and -1.34590 values between -1.34 and -0.502818 values between -0.50 and 0.35310 values between 0.35 and 1.1959 values between 1.19 and 2.0326 values between 2.03 and 2.8714 values between 2.87 and 3.717 values between 3.71 and 4.555 values between 4.55 and 5.393 values between 5.39 and 6.232 values between 6.23 and 7.083 values between 7.08 and 7.921 values between 7.92 and 8.762 values between 10.44 and 11.281 values between 11.28 and 12.122 values between 12.12 and 12.962 values between 12.96 and 13.81offnegativepositiveoften

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. cott
  2. 宠
  3. resco
  4. liệt
  5. cript
  6. egt
  7. sections
  8. fter
  9. pired
  10. ool

Suppresses

  1. balance
  2. Balance
  3. balance
  4. balances
  5. 平衡
  6. Balance
  7. _balance
  8. balancing
  9. Bal
  10. balanced