Neuron Explorer

Qwen3-0.6B

the digit 1 in COVID-19

A neuron in layer 2 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 number 1 in 'COVID-19', the one spot in the name where the model has to work out the rest of the word. Turning it up did not make the model's replies about COVID.

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

    to eggs.⏎Note, however, that the flu vaccine will not protect against COVID-19, and the COVID-19 vaccine will not protect against flu. Moreover

  2. #2 peak

    during the pandemic. Organizations and prescribers are under even greater pressure to help ensure patients adhere to their medication regimen. Many hospitals are overwhelmed with treating COVID-19 patients. If a medication nonadherence event requires a patient to receive emergency

  3. #3 peak

    countries and we expect it to soon be the dominant COVID-19 strain circulating worldwide.⏎“The world is watching in real-time as the COVID-

  4. #4 peak

    should be well informed about the disease-causing virus and the ways through which it spreads the disease. The only easy and effective way to protect yourself from COVID-19 is by using an alcohol-based massage or washing your hands frequently, and not

  5. #5 peak

    even more hesitant to leave their homes to go to the pharmacy. But it is particularly important that such higher-risk patients maintain adherence to reduce their vulnerability to COVID-19.⏎- Public mail carriers will typically leave packages in a mailbox or at a

  6. #6 peak

    related to COVID-19.⏎A major social determinant worsening these health inequities for patients with disabilities

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 a small town in a small town in a small town in a small town in a small town in a small town. The number of people in the small town is 100. The number of people in the small

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

    974 .. to describe a global TCP/IP network. So, you could also argue the Internet (<|endoftext|>このトピックには0件の返信が含まれ、1人の参加者がいます。2020年7月6

  2. #2 peak

    (var1 x)⏎> that changed, and there may be nothing about var11 . var1 $ x that⏎> changed, etc. Yet one is apparently forced to keep track of and⏎> reconstruct all those elements.⏎The compiler

  3. #3 peak

    vaccinations for workers and patients⏎04 - Measures to prevent central venous catheter-related infections⏎05 - Measures to prevent ventilator-associated pneumonia (nosocomial pneumonia)⏎06 - Measures to prevent surgical site infections⏎07 -

  4. #4 peak

    the state is one of the nation’s leading producers of biofuels, pumping out over 1.2 billion gallons per year.3⏎However, all is not well on the policy front. Indiana also has a renewable portfolio standard, but

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. L1:2670 suppresses
  2. L1:3043 weakly excites
  3. L0:2747 weakly suppresses
  4. L1:1104 weakly excites
  5. L0:557 weakly excites
  6. L0:469 weakly suppresses
  7. L0:2730 weakly excites
  8. L1:2922 weakly excites
  9. L1:3046 weakly excites
  10. L1:146 weakly excites
  11. L0:2690 weakly suppresses
  12. L0:1034 weakly excites
  13. L0:1091 weakly suppresses
  14. L1:198 weakly suppresses
  15. L1:2484 weakly suppresses
  16. L1:2620 weakly suppresses
  17. L0:2940 weakly excites
  18. L1:1912 weakly excites
  19. L0:2418 weakly excites
  20. L1:549 weakly excites "g" in "e.g." examples
  21. L0:2811 weakly excites
  22. L0:338 weakly excites
  23. L1:2240 weakly suppresses
  24. L1:1238 weakly excites

Feeds later layers

  1. L4:1242 suppresses
  2. L7:767 weakly suppresses
  3. L13:1166 weakly suppresses
  4. L7:2132 weakly suppresses
  5. L16:1072 weakly suppresses
  6. L20:49 weakly suppresses
  7. L17:1789 weakly suppresses
  8. L15:2100 weakly excites
  9. L14:2012 weakly excites
  10. L5:286 weakly excites
  11. L13:958 weakly suppresses
  12. L26:237 weakly suppresses
  13. L6:2935 weakly excites
  14. L15:1804 weakly excites
  15. L5:2988 weakly suppresses
  16. L24:340 weakly excites
  17. L23:643 weakly excites
  18. L9:1069 weakly excites
  19. L7:1958 weakly suppresses
  20. L21:14 weakly suppresses
  21. L3:1375 weakly excites
  22. L16:2107 weakly suppresses
  23. L22:1044 weakly excites
  24. L7:249 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.

4 values between -0.95 and -0.845 values between -0.84 and -0.737 values between -0.73 and -0.618 values between -0.61 and -0.5011 values between -0.50 and -0.3937 values between -0.39 and -0.27137 values between -0.27 and -0.16858 values between -0.16 and -0.051905 values between -0.05 and 0.07720 values between 0.07 and 0.18179 values between 0.18 and 0.3027 values between 0.30 and 0.412 values between 0.41 and 0.524 values between 0.52 and 0.641 values between 0.86 and 0.981 values between 2.57 and 2.681 values between 3.25 and 3.361 values between 3.48 and 3.59offnegativepositiveoften

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. shit
  2. osphate
  3. },{"
  4. urse
  5. ัส
  6. cử
  7. Duterte
  8. js
  9. souvenir
  10. 放弃了

Suppresses

  1. 界面
  2. neys
  3. omid
  4. ney
  5. sthrough
  6. linked
  7. competitive
  8. ouver
  9. ekte
  10. oup