Neuron Explorer

Qwen3-0.6B

laboratory and research facility terms

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 responds to words about labs and research work, such as fume hoods, vacuum chambers, safety cabinets, and supercomputing. When it was turned up, the model's writing drifted toward lab tasks like syringes, plasma samples and experiments.

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

    interfaces and even 3D analysis of a sample. It is a very sensitive technique that requires a very high vacuum (10-9~10-11 mbar). As schem

  2. #2 peak

    to keep unnecessary equipment and glassware outside of the hood at all times and store all chemicals in approved storage cans, containers, or cabinets (not in the fume hood). Furthermore, it is best to keep the work space neat and clean in

  3. #3 peak

    and check it every time the hood is used.⏎(For further information, see section 8.C.5 on testing and verification.)⏎Laboratory fume hoods and adjacent work areas should be kept clean and free of debris at all times.

  4. #4 peak

    on. Most major cities have a supplier that will sell it to you, but very often they have large minimum orders and/or require that you own an expensive dewar flask into which they may safely dispense the liquid nitrogen. At -19

  5. #5 peak

    potentially help India leapfrog ahead in the supercomputing world. Its high-performance, energy efficient NVIDIA GPU accelerators will enable Indian researchers to build next-generation supercomputing technologies while maximising costs and power savings - a critical challenge to overcome for

  6. #6 peak

    , theoretical or computational methods6,7.⏎In order to provide a bottom-up description of materials behavior, our interdisciplinary team of undergraduate students, graduate students and postdocs applies an experimentally validated atomistic based multi-scale simulation approach that considers the structure

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 to read the news and see the something that is related to the question and answer: I have to read the statement "I am a person and the something that is related to the question and answer" the statement "I am a

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

    fail more, and so currently there are not enough organs to go around. In fact, in the last 10 years, the number of patients requiring an organ has doubled, while in the same time, the actual number of transplants has

  2. #2 peak

    anced by the National Centre for Research and Development under grant No. SP/I/1/77065/10 by the strategic scientific research and experimental development program:⏎SYNAT - “Interdisciplinary System for Interactive Scientific and Scientific

  3. #3 peak

    1955 to 1957 he was a member of the USAF precision flying team, the Thunderbirds. He later was an assistant professor in the mathematics department at the Air Force Academy, and in 196

  4. #4 peak

    Really, no one is," says kidney donor Krystal McLear, 32, who serves on a committee for the network that runs the U.S. organ transplant system.⏎The United Network for Organ Sharing is debating some new policies to change

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. L3:1240 suppresses
  2. L3:1276 suppresses
  3. L6:2815 suppresses
  4. L6:2165 excites
  5. L6:1772 suppresses
  6. L3:185 excites
  7. L5:2474 suppresses
  8. L3:105 excites
  9. L6:658 excites
  10. L5:1146 excites
  11. L6:2130 excites
  12. L4:1946 suppresses
  13. L6:453 excites
  14. L6:2758 weakly suppresses
  15. L6:17 weakly suppresses
  16. L6:827 weakly suppresses
  17. L4:2316 weakly suppresses
  18. L0:2532 weakly suppresses
  19. L6:1375 weakly excites
  20. L4:2788 weakly suppresses
  21. L6:2128 weakly suppresses
  22. L6:1129 weakly suppresses
  23. L4:2531 weakly excites
  24. L5:132 weakly excites

Feeds later layers

  1. L25:1338 excites
  2. L21:206 suppresses
  3. L21:1105 suppresses
  4. L11:527 excites
  5. L11:1489 suppresses
  6. L10:656 excites
  7. L24:131 excites
  8. L21:938 excites
  9. L24:2986 excites
  10. L8:923 excites
  11. L18:1494 excites
  12. L21:1870 excites
  13. L23:2744 excites
  14. L24:1944 excites
  15. L9:179 suppresses
  16. L22:2340 excites
  17. L24:2988 excites
  18. L22:509 excites
  19. L9:2386 excites
  20. L17:1142 excites
  21. L25:949 excites
  22. L23:1582 excites
  23. L26:2344 excites
  24. L23:2684 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.

1 values between -3.97 and -3.771 values between -2.57 and -2.371 values between -2.17 and -1.971 values between -1.97 and -1.772 values between -1.57 and -1.373 values between -1.37 and -1.175 values between -0.97 and -0.7711 values between -0.77 and -0.57119 values between -0.57 and -0.37832 values between -0.37 and -0.172206 values between -0.17 and 0.03585 values between 0.03 and 0.2373 values between 0.23 and 0.4322 values between 0.43 and 0.6320 values between 0.63 and 0.836 values between 0.83 and 1.036 values between 1.03 and 1.234 values between 1.23 and 1.432 values between 1.43 and 1.634 values between 1.63 and 1.831 values between 1.83 and 2.031 values between 3.23 and 3.432 values between 3.83 and 4.03offnegativepositiveoften

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. 研究人员
  2. 科研
  3. 醮
  4. steril
  5. 师兄
  6. ctica
  7. microscopy
  8. PhD
  9. patient
  10. 汞

Suppresses

  1. 联社
  2. Commerce
  3. Social
  4. _social
  5. earning
  6. .head
  7. 部落
  8. 消费品
  9. social
  10. ⏎