Neuron Explorer

Qwen3-0.6B

'study' in science news

A neuron in layer 12 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 'study' (or 'new study') when a news article or abstract is talking about a research study. Turning it up gave text that repeated phrases or drifted, so the steering does not clearly show the same idea.

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

    Voltages and Currents⏎An alternating voltage may be generated⏎(1) by rotating a coil at constant angular velocity in a uniform magnetic field as shown in Fig. (a).⏎(2) by rotating a magnetic field at a constant angular

  2. #2 peak

    . This can make the bonding process more difficult. Finding ways to help parents and these infants bond better is important for everyone in the family.⏎The results of this study on reading has been published in the journal Journal of Developmental and Behavioral Pediatrics.

  3. #3 peak

    hippocampus, which is a part of the brain responsible for much of our memory and learning functions. Our hippocampal volume naturally decreases with age, but this study suggests that regular consumption of sugar-sweetened beverages may accelerate this process.⏎D

  4. #4 peak

    found in fermented foods<|endoftext|>Two plasma biomarkers were notably associated with when athletes return to action after concussions, according to a new study of collegiate athletes and recovery time. “Although preliminary, the current results highlight

  5. #5 peak

    became available can now be applied much more widely on a whole range of conservation issues,” said WCS Conservation Support scientist Dr. Robert Rose, the lead author of the study. “To that end, we sought out the top thought leaders in conservation and

  6. #6 peak

    being consumed.)⏎A 12-ounce can of soda typically harbors about 150 sugary calories (which scientists, including the authors of the new study, confusingly call kilocalories). Many candy bars contain more calories than

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: A, B, C, D, and, E, F, G, H, I, and, J, and, J, and, J, and, K, and, J, and, J, and, J,

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

    of plastic waste in oceans has detrimental effects on marine wildlife. The following infographics help visualize trends in waste generation, including plastics, for 192 coastal countries. Data is based on the works of Jambeck et al. (20

  2. #2 peak

    HOTBOX only 5 times per week your household would save 119kg of CO2 per year. If 500 000 households did that it amounts to more than 60 000 tons of

  3. #3 peak

    have. The USGS operates programs that include natural hazards research, such as our earthquake, volcano, and landslide programs; a network of 8,000 streamgages that monitor water availability and help in forecasting floods; and other programs that

  4. #4 peak

    Neglect, Adoption & Foster Care⏎- National Survey of Child and Adolescent Well-Being (NSCAW), 1997-2014 | Learn more about this project⏎- NSCAW: Research Briefs⏎

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. L11:2791 excites
  2. L7:1057 excites
  3. L9:570 weakly excites
  4. L4:63 weakly suppresses
  5. L10:140 weakly suppresses
  6. L8:1577 weakly excites
  7. L8:429 weakly excites
  8. L11:965 weakly excites
  9. L7:1521 weakly excites
  10. L9:541 weakly excites
  11. L8:2274 weakly excites
  12. L11:1190 weakly suppresses
  13. L11:54 weakly excites
  14. L1:1263 weakly suppresses
  15. L11:3029 weakly excites
  16. L10:1342 weakly suppresses
  17. L11:2661 weakly excites
  18. L10:158 weakly excites
  19. L9:2180 weakly excites
  20. L8:2009 weakly excites
  21. L6:2070 weakly suppresses
  22. L7:1835 weakly suppresses
  23. L11:431 weakly excites
  24. L1:118 weakly suppresses

Feeds later layers

  1. L13:1648 suppresses
  2. L13:2227 suppresses
  3. L16:1748 weakly excites
  4. L13:676 weakly excites
  5. L13:1936 weakly suppresses
  6. L17:2944 weakly suppresses
  7. L15:2589 weakly suppresses
  8. L25:652 weakly excites
  9. L19:2235 weakly suppresses
  10. L18:3016 weakly suppresses
  11. L20:3057 weakly suppresses "of" after "people" or "women"
  12. L19:1171 weakly suppresses
  13. L15:2322 weakly suppresses
  14. L16:96 weakly excites
  15. L16:2159 weakly suppresses
  16. L26:929 weakly suppresses
  17. L16:2330 weakly excites
  18. L16:1749 weakly suppresses
  19. L17:2457 weakly suppresses
  20. L16:712 weakly suppresses
  21. L19:1781 weakly excites
  22. L15:2392 weakly suppresses
  23. L23:3017 weakly suppresses
  24. L14:685 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.

2 values between -2.07 and -1.941 values between -1.56 and -1.443 values between -1.44 and -1.316 values between -1.31 and -1.187 values between -1.18 and -1.0616 values between -1.06 and -0.9327 values between -0.93 and -0.8053 values between -0.80 and -0.6871 values between -0.68 and -0.55149 values between -0.55 and -0.42286 values between -0.42 and -0.30529 values between -0.30 and -0.17881 values between -0.17 and -0.041159 values between -0.04 and 0.09321 values between 0.09 and 0.21169 values between 0.21 and 0.3467 values between 0.34 and 0.4743 values between 0.47 and 0.5934 values between 0.59 and 0.7226 values between 0.72 and 0.8515 values between 0.85 and 0.9714 values between 0.97 and 1.104 values between 1.10 and 1.233 values between 1.23 and 1.355 values between 1.35 and 1.484 values between 1.48 and 1.612 values between 1.61 and 1.732 values between 1.73 and 1.861 values between 1.86 and 1.991 values between 1.99 and 2.111 values between 2.11 and 2.241 values between 2.24 and 2.372 values between 2.37 and 2.491 values between 2.62 and 2.751 values between 2.75 and 2.871 values between 2.87 and 3.00offnegativepositiveoften

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. 嗡
  2. ffi
  3. 材
  4. Background
  5. ff
  6. 的文字
  7. ulation
  8. Permanent
  9. vain
  10. /rec

Suppresses

  1. UCH
  2. ouch
  3. чи
  4. rito
  5. ByUsername
  6. controversies
  7. жи
  8. nda
  9. uch
  10. 手