Neuron Explorer

Qwen3-0.6B

unclear: 'Match' in exercise instructions

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 a few spots in homework and exercise text, like the word 'Match' that starts an instruction or a symbol in a formula. It turns off on words such as 'relationships'. Turning it up or down gave no clear, consistent topic.

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

    C, higher concentration less lower concentration) divided by the thickness of the membrane. At steady state, Fickian diffusion can be approximated by:⏎J = D (CWaterlooEmitter - groundwater) / membrane wall thickness⏎where D is

  2. #2 peak

    ? What is stopping both homologues from going into one daughter cell?⏎8. If there is some force holding homologous chromosomes together - when a pair of homologous chromosomes line up on the equ

  3. #3 peak

    has patented the most inventions, the most illuminating being the incandescent electric bulb, the phonograph and the motion-picture projector. Who was he?⏎- Match the following with their inventors:⏎(a

  4. #4 peak

    . There are 4 different tasks which include:⏎~Fill in the missing numbers⏎~Match line to equation⏎~Create line using an equation⏎~Create equation using a

  5. #5 peak

    are separated in Anaphase 2 of meiosis, does that mean that 2 of the resulting gametes are genetically identical to each other [provided crossing-over hasn't taken place

  6. #6 peak

    , owners, and other providers of capital, or they will invest elsewhere.⏎The following is the WACC calculation formula:⏎WACC = E/V × Re + D/V × Rd × (1 - Tc)⏎Re = cost of equity⏎

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 relationship between the three elements of the same kind in a single object or relationship of existence in a set. For example, "A man and a woman of relationship in the living world." For example, "The relationship between the three

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

    healthier an area is, the more biodiversity it has. Different forms of wildlife and plants inhabit areas, and these plants and animals learn to coexist and form ecological relationships with each other.⏎In unhealthy environments, only a few kinds of species of each

  2. #2 peak

    emphasized the concepts of morality and ritualism. The term “ritual propriety” (禮, pinyin: lǐ), describes the “proper” social relationships and the set of rituals which regulate them. The fundamental social relationships are those between

  3. #3 peak

    nature of reality, but primarily predicated on the the fact that humans subjectively interpret the world in terms of objects (vertices) and their respective relationships to one another (edges) (an argument against this idea). The popular data model used in

  4. #4 peak

    ), focuses on changing negative content of thoughts while MBCT emphasizes the process of paying attention to thoughts and feelings moment by moment and without judgment. Changing the patient's relationship to the suffering caused by negative thoughts is the key because there is no possible way

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. L6:2295 suppresses
  2. L6:1239 suppresses
  3. L6:1284 suppresses
  4. L2:2676 weakly excites
  5. L5:1643 weakly suppresses
  6. L6:2160 weakly excites
  7. L5:1941 weakly suppresses
  8. L1:2722 weakly suppresses
  9. L6:1396 weakly suppresses
  10. L6:830 weakly excites
  11. L6:1185 weakly excites
  12. L2:2962 weakly suppresses
  13. L5:1379 weakly excites
  14. L6:1791 weakly excites
  15. L0:2109 weakly excites
  16. L4:1483 weakly excites
  17. L4:2342 weakly suppresses
  18. L5:1140 weakly suppresses 'of' after a page, article or section number
  19. L3:272 weakly excites
  20. L3:1944 weakly suppresses
  21. L4:21 weakly excites
  22. L1:2374 weakly excites
  23. L5:102 weakly excites
  24. L5:28 weakly excites

Feeds later layers

  1. L9:149 suppresses
  2. L22:1760 suppresses
  3. L24:741 suppresses
  4. L17:3040 suppresses
  5. L9:3068 suppresses
  6. L11:1345 suppresses
  7. L10:908 suppresses
  8. L16:1818 suppresses "two" and the link between two things
  9. L11:1157 weakly suppresses
  10. L10:1418 weakly suppresses
  11. L11:2972 weakly suppresses
  12. L16:2226 weakly suppresses
  13. L22:1257 weakly suppresses
  14. L14:290 weakly suppresses
  15. L18:2294 weakly suppresses
  16. L18:2373 weakly suppresses
  17. L10:957 weakly suppresses
  18. L10:883 weakly suppresses
  19. L22:2920 weakly excites
  20. L8:1741 weakly suppresses
  21. L10:2767 weakly excites
  22. L8:2436 weakly suppresses
  23. L21:1782 weakly suppresses
  24. L25:2524 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 -4.69 and -4.551 values between -4.12 and -3.972 values between -1.96 and -1.811 values between -1.81 and -1.672 values between -1.67 and -1.521 values between -1.52 and -1.382 values between -1.38 and -1.241 values between -0.95 and -0.805 values between -0.80 and -0.663 values between -0.66 and -0.5218 values between -0.52 and -0.37117 values between -0.37 and -0.23571 values between -0.23 and -0.081884 values between -0.08 and 0.061037 values between 0.06 and 0.20220 values between 0.20 and 0.3534 values between 0.35 and 0.493 values between 0.49 and 0.642 values between 0.64 and 0.781 values between 0.78 and 0.922 values between 0.92 and 1.07offnegativepositiveoften

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. amping
  2. ạt
  3. opus
  4. amped
  5. essional
  6. 政策
  7. _flashdata
  8. amnesty
  9. insula
  10. Memories

Suppresses

  1. 关系
  2. 合作关系
  3. interacting
  4. 协调
  5. 相处
  6. 交往
  7. interpersonal
  8. 自然界
  9. relationship
  10. 干涉