Neuron Explorer

Qwen3-0.6B

cooking and canning appliances

A neuron in layer 7 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 fires on words about cooking and preserving food, such as 'pressure cooker', 'pasteurization', 'slow cookers' and 'canning', and also on a few 'volcanism' words. It turns off on words meaning 'own' or 'themselves', like 'their own' or 'for themselves'. Steering changed the model's topic, but not in a clearly food related direction.

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

    use only as much flame as necessary (Remember that slightly boiling water cooks foods at the same temperature as water that is rapidly boiling).⏎- Pressure cooker: use a pressure cooker more often since it cooks food fast over low heat.⏎- Cook in one

  2. #2 peak

    from USDA. For many of these tomato products, there are canning options for both boiling water and pressure canning available in our directions.⏎In the case of these tomato products with both

  3. #3 peak

    potentially harmful bacteria (like salmonella and E. coli 0157:H7 – the most dangerous kind of E.coli) in milk. Pasteurization also increases the shelf-life of milk because the bacteria in raw milk make it

  4. #4 peak

    a traditional gas or electric stove.⏎Consistently using these small appliances can make a huge difference in your energy consumption, Newgent said. Even though slow cookers are typically on for hours at a time, they will burn less energy than a

  5. #5 peak

    cast day, because you need heat and direct sunlight to cook the s’mores. You will need graham crackers, marshmallows, chocolate bars, aluminum foil, plastic wrap, tape, glue sticks, and cardboard boxes with lids. T

  6. #6 peak

    earthquake ever recorded on the Earth at magnitude 9.5. The most recent (confirmed) eruption of the Puyehue volcano according to the Global Volcanism Program was in AD 860 (+/-75 years).⏎The

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 first time, I have been to the city, and I have been to the river. This is a problem. This is a solution. The problem. This is a solution. The problem. This is a solution. The problem.

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

    tradition of beginning physics with careful measurements of volumes and masses and calculations of densities...⏎This activity allows pupils to make some measurements to measure the density of a material for themselves.⏎To appreciate the idea of density just imagine you can look deep inside a material

  2. #2 peak

    the result of selective communication from scientists and a media that too often have been more than eager to promote these stories without engaging in any critical analysis or research of their own.

  3. #3 peak

    program.⏎In order to keep India from falling to communism, America institutes a⏎program of food and monetary aid. This had the effect of reducing⏎India's own production of wheat and other food as they could

  4. #4 peak

    , make sure you set aside enough time to consider them fully and, if appropriate, plan your response⏎- Try and avoid looking only at the detail of your own intervention and getting too bogged down in the demands flowing from it⏎- Keep

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. L4:1407 suppresses
  2. L4:2453 suppresses
  3. L3:2402 excites
  4. L6:2713 suppresses
  5. L4:1699 suppresses
  6. L6:2990 excites
  7. L2:2502 weakly excites
  8. L3:1873 weakly suppresses
  9. L4:1805 weakly excites
  10. L6:2453 weakly suppresses
  11. L5:1892 weakly excites
  12. L6:1123 weakly suppresses
  13. L5:671 weakly excites
  14. L2:856 weakly excites
  15. L3:2623 weakly excites
  16. L6:2559 weakly excites
  17. L3:2275 weakly suppresses
  18. L6:1047 weakly excites
  19. L2:1754 weakly excites
  20. L5:2780 weakly excites
  21. L4:2835 weakly suppresses semicolons joining two clauses
  22. L6:209 weakly excites
  23. L4:2936 weakly suppresses
  24. L2:408 weakly suppresses

Feeds later layers

  1. L9:1437 suppresses
  2. L17:1485 suppresses
  3. L13:1598 suppresses
  4. L10:223 suppresses
  5. L13:1930 suppresses
  6. L21:1627 suppresses
  7. L18:2510 suppresses
  8. L19:3014 suppresses
  9. L10:2979 suppresses
  10. L10:2853 weakly suppresses
  11. L11:2060 weakly suppresses
  12. L9:1200 weakly suppresses
  13. L11:107 weakly excites
  14. L15:2530 weakly suppresses
  15. L10:2788 weakly suppresses
  16. L10:1708 weakly suppresses
  17. L12:597 weakly suppresses
  18. L13:2117 weakly excites
  19. L13:1984 weakly suppresses
  20. L13:453 weakly suppresses
  21. L22:2305 weakly suppresses
  22. L25:849 weakly suppresses
  23. L14:688 weakly suppresses
  24. L8:400 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 -7.28 and -7.051 values between -6.35 and -6.111 values between -4.94 and -4.711 values between -3.77 and -3.541 values between -2.84 and -2.601 values between -2.60 and -2.371 values between -2.37 and -2.141 values between -2.14 and -1.902 values between -1.90 and -1.671 values between -1.67 and -1.435 values between -1.20 and -0.972 values between -0.97 and -0.7311 values between -0.73 and -0.5078 values between -0.50 and -0.261088 values between -0.26 and -0.032320 values between -0.03 and 0.21358 values between 0.21 and 0.4427 values between 0.44 and 0.674 values between 0.67 and 0.911 values between 0.91 and 1.141 values between 1.37 and 1.611 values between 1.61 and 1.841 values between 1.84 and 2.08offnegativepositiveoften

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. ugin
  2. alm
  3. ram
  4. ramento
  5. peg
  6. ritis
  7. 国际机场
  8. relation
  9. otion
  10. rams

Suppresses

  1. own
  2. 自己的
  3. dden
  4. собой
  5. OWN
  6. Own
  7. włas
  8. 了自己的
  9. IBUTE
  10. własne