Neuron Explorer

Qwen3-0.6B

'the' in 'on the other hand' phrases

A neuron in layer 20 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 'the' in set phrases like 'on the other hand', 'on the one hand' and 'on the right'. When it was turned up, the model got stuck repeating the word 'one', which is the word the neuron seems to push toward.

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

    25 times more potent than carbon dioxide.⏎In the Communication, the Commission recommends realizing bio-waste potential as a renewable source of energy and recycled materials, on the basis of enhanced implementation of existing legislation at the national level with support at the EU

  2. #2 peak

    on hands-free mobiles while driving⏎Brake has renewed its call for the government to look again at the law surrounding hands-free mobile phones and driving, on the back of new research which concluded that talking on a hands-free phone can be just

  3. #3 peak

    child's development and, on the 'Words for Life' website, there are some useful tips about how you can

  4. #4 peak

    to be made. A pop icon or populist president might discover advantages in gainsaying the grammar of standard English in favour of a local dialect or patois. On the other hand, scrupulous attention to the rules of prescriptive grammar might be

  5. #5 peak

    weekly earnings by worker characteristics, including educational attainment on the right-hand side:⏎Source: New York Times, Bureau of Labor Statistics⏎For

  6. #6 peak

    genotypes may represent an important source of genes, with which in the future breeding programs can create varieties with high adaptability to adverse impacts of climate change. On the other hand, local collection of these genotypes will contribute to the identification, protection

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 fact that, while most of the world is focused in the process of our own process of our own own own own of our own of our own of our own of our own of our own of our own of our own of our own

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

    non-torpor Mars mission approaches and established the medical plausibility of torpor. We have expanded our team and put together a development plan that we are in the process of executing.⏎SpaceWorks makes clear that "enabling access to Mars is

  2. #2 peak

    regions, and local governments have started thinking about their own Green New Deals (GNDs) to address this crisis now. Municipalities and localities are in the unique position to enact truly transformative climate justice policies due to their ability to be much

  3. #3 peak

    a cobot is?⏎You might have come across this term “cobots” while searching for a solution to improve your production processes but weren’t quite sure what the term referred to. As you’ve probably noticed yourself, cobots are trending more

  4. #4 peak

    by the War Department, and continues to the present day, despite military public health efforts to reduce tobacco use. Military recruits, often young and away from home for the first time,

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. L17:30 suppresses
  2. L16:792 excites "the" inside a prepositional phrase
  3. L9:24 excites "the" before a named organisation or title
  4. L19:427 suppresses
  5. L4:52 excites
  6. L10:120 suppresses
  7. L16:419 excites
  8. L18:641 suppresses
  9. L2:205 weakly excites
  10. L7:533 weakly suppresses
  11. L13:975 weakly excites
  12. L18:811 weakly excites
  13. L15:1848 weakly suppresses
  14. L11:1168 weakly excites
  15. L18:171 weakly excites
  16. L3:291 weakly excites
  17. L8:636 weakly suppresses
  18. L9:662 weakly suppresses
  19. L15:2320 weakly suppresses
  20. L16:1293 weakly suppresses
  21. L10:1377 weakly suppresses
  22. L19:2283 weakly suppresses
  23. L12:435 weakly suppresses
  24. L16:1949 weakly suppresses

Feeds later layers

  1. L21:1237 excites
  2. L26:1028 excites
  3. L27:556 excites
  4. L24:979 weakly excites
  5. L21:2976 weakly excites
  6. L24:823 weakly excites
  7. L23:1004 weakly excites
  8. L25:1926 weakly excites
  9. L25:2369 weakly excites
  10. L25:572 weakly excites
  11. L24:2713 weakly excites 'a' before few or variety
  12. L24:1552 weakly excites
  13. L25:1990 weakly excites
  14. L25:2165 weakly excites
  15. L23:742 weakly suppresses
  16. L25:549 weakly suppresses
  17. L26:2620 weakly excites
  18. L23:2525 weakly excites
  19. L23:1361 weakly excites
  20. L24:2538 weakly excites
  21. L21:1489 weakly excites
  22. L27:2548 weakly suppresses
  23. L23:1633 weakly excites
  24. L26:1432 weakly 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 -14.89 and -14.181 values between -14.18 and -13.461 values between -12.74 and -12.031 values between -12.03 and -11.312 values between -11.31 and -10.601 values between -9.16 and -8.452 values between -8.45 and -7.734 values between -7.73 and -7.014 values between -7.01 and -6.304 values between -6.30 and -5.5811 values between -5.58 and -4.878 values between -4.87 and -4.1510 values between -4.15 and -3.4310 values between -3.43 and -2.7221 values between -2.72 and -2.0020 values between -2.00 and -1.2964 values between -1.29 and -0.572705 values between -0.57 and 0.15929 values between 0.15 and 0.8638 values between 0.86 and 1.5827 values between 1.58 and 2.2915 values between 2.29 and 3.017 values between 3.01 and 3.732 values between 3.73 and 4.445 values between 4.44 and 5.164 values between 5.16 and 5.873 values between 5.87 and 6.591 values between 8.02 and 8.741 values between 8.74 and 9.452 values between 9.45 and 10.171 values between 10.17 and 10.892 values between 11.60 and 12.321 values between 13.03 and 13.75offnegativepositiveoften

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. 一页
  2. one
  3. one
  4. 一步
  5. 一位
  6. 一个人
  7. 一种
  8. 一站
  9. 一个问题
  10. satu

Suppresses

  1. -final
  2. isateur
  3. LEMENT
  4. tract
  5. msm
  6. heck
  7. lement
  8. .bn
  9. ulant
  10. anye