Neuron Explorer

Qwen3-0.6B

"raw" as in raw materials

A neuron in layer 5 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 word "raw," especially in phrases like "raw materials" for industry. It is mostly quiet on other words. Turning it up made the model's writing wander without a clear link to materials or industry.

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

    product quality have driven exponential change in the U.S. steel industry. Steel-making in the United States has traditionally been dominated by integrated steel mills: mills that use raw materials to produce steel products. Since the 1980s, however

  2. #2 peak

    areas where rice is difficult to grow. In those areas, rice, millet, sorghum, and corn are the staples. Rice is accompanied by a raw salad of leaves, fruit, or vegetables; a soup; and curries of

  3. #3 peak

    , economic and scientific and technical cooperation (1993). In Vietnam, Ukrainian enterprises export grain, honey, nuts, plant seeds, sunflower oil, raw materials for the light and chemical industries, rolled metal products, and antibiotics. Sp

  4. #4 peak

    word meaning ‘hemp‘. All of the British captains carried hemp seed in their stores to be sewn in any new lands they visited. This would provide raw materials for future explorers for when they needed new sails, ropes etc<|endoftext|>King

  5. #5 peak

    important to develop it continuously.⏎4. Standardisation and Grading:⏎Standardisation refers to determining of standard regarding size, quality, design, weight, colour, raw material to be used, etc., in respect of a particular product. By doing

  6. #6 peak

    .⏎Logistics vs. supply chain⏎These are similar terms, so let’s discuss how they are different. A supply chain refers to the entire process of converting raw material to a finished product and delivering it to the customer. Logistics is the detailed

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: in the context of the United States, in the past, and in the present, and in the work of the United States Government, including the Department of the Treasury, and in the United States, including the President’s Office of Foreign and

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

    links to you. To join WhatsApp groups click the below links.⏎MUST JOIN VU STUDY GROUPS⏎CS610: Computer Networks⏎A network is defined as a system for connecting computers using a single transmission technology. Computers can

  2. #2 peak

    a thorough scan on your system. Before you do that, though, take a few extra precautions.⏎Is your computer part of a network?⏎If so, you'll need to disconnect from it immediately. That's to

  3. #3 peak

    the basic technical concepts of Computer Networks. Be familiar with the various types of network configurations. Describe the circuits that are available for voice and data networks, their transmission

  4. #4 peak

    blockage of the carotid and vertebral arteries with fatty deposits, which leads to reduced blood flow to the brain, and then stroke.⏎When a blood clot blocks a blood vessel in the brain or neck, it is called a thrombus

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. L2:2730 suppresses
  2. L3:1006 weakly excites
  3. L4:1899 weakly excites
  4. L4:2450 weakly suppresses
  5. L3:2133 weakly suppresses
  6. L3:1437 weakly suppresses
  7. L0:656 weakly suppresses 'to' after 'as opposed' or 'related'
  8. L2:2696 weakly excites
  9. L2:541 weakly suppresses
  10. L3:1037 weakly suppresses
  11. L2:2317 weakly excites
  12. L1:1000 weakly suppresses
  13. L1:720 weakly excites
  14. L2:2912 weakly suppresses
  15. L4:1549 weakly suppresses
  16. L2:1660 weakly excites
  17. L0:2949 weakly excites
  18. L2:1736 weakly excites
  19. L1:1385 weakly suppresses
  20. L4:1041 weakly suppresses
  21. L3:2958 weakly excites
  22. L2:2765 weakly suppresses
  23. L4:1299 weakly suppresses
  24. L4:2956 weakly suppresses

Feeds later layers

  1. L9:755 weakly excites
  2. L6:734 weakly suppresses
  3. L6:184 weakly excites
  4. L15:2213 weakly suppresses
  5. L10:93 weakly excites
  6. L16:2117 weakly excites
  7. L12:169 weakly excites
  8. L26:1978 weakly suppresses
  9. L10:1048 weakly excites
  10. L27:1756 weakly excites
  11. L11:2004 weakly excites
  12. L25:2060 weakly excites
  13. L7:1038 weakly suppresses
  14. L7:646 weakly excites bodily samples and medical tests
  15. L12:2467 weakly suppresses
  16. L7:571 weakly suppresses
  17. L6:2091 weakly suppresses
  18. L7:2863 weakly suppresses
  19. L12:2011 weakly suppresses
  20. L11:340 weakly suppresses
  21. L10:1178 weakly excites
  22. L16:744 weakly excites
  23. L25:990 weakly suppresses
  24. L16:1329 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 -0.96 and -0.903 values between -0.90 and -0.841 values between -0.78 and -0.721 values between -0.72 and -0.662 values between -0.66 and -0.603 values between -0.60 and -0.542 values between -0.54 and -0.482 values between -0.48 and -0.4211 values between -0.42 and -0.3632 values between -0.36 and -0.3069 values between -0.30 and -0.24139 values between -0.24 and -0.18281 values between -0.18 and -0.12463 values between -0.12 and -0.06884 values between -0.06 and -0.00984 values between -0.00 and 0.06476 values between 0.06 and 0.12267 values between 0.12 and 0.18137 values between 0.18 and 0.2478 values between 0.24 and 0.3039 values between 0.30 and 0.367 values between 0.36 and 0.428 values between 0.42 and 0.483 values between 0.48 and 0.541 values between 0.54 and 0.603 values between 0.60 and 0.651 values between 0.65 and 0.711 values between 0.71 and 0.772 values between 0.77 and 0.832 values between 0.83 and 0.891 values between 0.89 and 0.951 values between 0.95 and 1.011 values between 1.01 and 1.071 values between 1.13 and 1.191 values between 1.37 and 1.43offnegativepositiveoften

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. MLS
  2. isi
  3. ragments
  4. sti
  5. ?><?
  6. <Entry
  7. >Data
  8. 前线
  9. 膳
  10. passes

Suppresses

  1. worth
  2. beau
  3. ication
  4. navegador
  5. reck
  6. Navigation
  7. shake
  8. đen
  9. ัน
  10. coon