Neuron Explorer

Qwen3-0.6B

formal nouns after verbs of answering or committing

A neuron in layer 10 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 lights up on words that follow a verb of answering, reporting, committing or explaining, such as 'answers', 'report', 'explain' or 'committed'. When it was turned up, the text drifted into vague, bureaucratic phrasing, with no clear change in topic.

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

    critical, missing long-term framework for informing and coordinating policy and business decisions. Internationally, they provide a transparent benchmark for evaluating national commitments.⏎DDPP research teams study decarbonization in their own countries. Each team

  2. #2 peak

    workplace safety and health standards, encourages voluntary compliance through consultation and partnerships, and promotes safety training and education for workers and employers.

  3. #3 peak

    to report the number of animals euthanized and adopted each year and will penalize laboratories that do not comply with adoption standards. Currently, labs are required to report to the U.S. Department of Agriculture how many animals are used each year,

  4. #4 peak

    to correct any misbehavior on the part of the addressee. Ideally, the erring individual is given a warning letter to provide him with ample time to explain his side on the matter as part of the due process. This will give him

  5. #5 peak

    might point to a specific area on the page and ask her questions about it. For example:⏎“Can you tell me about what you did here?” Dana's response: “Car, vroom.”⏎Repeat and perhaps add to her comment. For

  6. #6 peak

    their bodies, but they ultimately relied on the slaves to play their part as valuable commodities. Slave buyers stripped the slaves and questioned their pasts, seeking more honest answers than they could get from the traders. In turn, these examinations provided information that

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 need to get this done, and I need to get to the next step. So I need to get to the next step, and I need to get to the next step, and I need to get to the next step, 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

    the well-being of the children. The Guardian ad Litem’s job is to act in the best interests of the child(ren) in reporting back to the Judge and making recommendations. The Guardian can obtain access to medical and other records

  2. #2 peak

    It’s actually not a bad idea,” Trump said. Kim answered through an interpreter, saying “I think that is something that could be welcomed.”⏎Sweden permitted a reporter

  3. #3 peak

    horizontal. (seaplane) trim-The angle with the horizontal surface of the water assumed by the float or hull under given conditions. trimming moment-The moment about a reference point, usually the center of gravity, exerted by the seaplane

  4. #4 peak

    he will utilize photos and video footage from NASA satellites to discuss the visual phenomenon known as the Northern Lights. This panel will be moderated by Joseph M. Davila, an Astrophysicist in the Solar Physics Branch

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. L3:1422 excites
  2. L7:1039 weakly excites
  3. L8:400 weakly excites
  4. L9:2698 weakly excites
  5. L9:2703 weakly excites
  6. L9:2407 weakly excites
  7. L9:2782 weakly excites
  8. L9:2679 weakly suppresses
  9. L7:1857 weakly excites
  10. L0:2346 weakly suppresses
  11. L8:2350 weakly excites
  12. L4:1940 weakly suppresses
  13. L5:1444 weakly suppresses
  14. L7:1028 weakly suppresses
  15. L8:1016 weakly suppresses
  16. L0:1058 weakly suppresses
  17. L5:160 weakly excites
  18. L5:2358 weakly excites
  19. L7:2119 weakly excites
  20. L5:86 weakly suppresses
  21. L7:940 weakly suppresses
  22. L9:2928 weakly suppresses
  23. L6:1085 weakly excites
  24. L8:1766 weakly suppresses

Feeds later layers

  1. L12:2994 suppresses
  2. L11:1259 suppresses coercion, abuse and harassment
  3. L11:2195 excites
  4. L17:1162 suppresses
  5. L12:2634 suppresses
  6. L12:762 excites
  7. L11:2278 suppresses
  8. L11:117 excites
  9. L18:1961 suppresses
  10. L12:767 excites
  11. L11:375 suppresses
  12. L11:2511 suppresses
  13. L11:2968 excites
  14. L11:356 suppresses
  15. L13:2960 weakly suppresses
  16. L11:745 weakly excites
  17. L11:2420 weakly excites
  18. L21:1473 weakly excites
  19. L16:2788 weakly suppresses
  20. L17:705 weakly suppresses
  21. L11:2125 weakly excites
  22. L11:1523 weakly suppresses
  23. L12:667 weakly suppresses
  24. L12:1923 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.

3 values between -1.71 and -1.593 values between -1.46 and -1.343 values between -1.34 and -1.227 values between -1.22 and -1.104 values between -1.10 and -0.985 values between -0.98 and -0.864 values between -0.86 and -0.7412 values between -0.74 and -0.6219 values between -0.62 and -0.5040 values between -0.50 and -0.3898 values between -0.38 and -0.25268 values between -0.25 and -0.131014 values between -0.13 and -0.011733 values between -0.01 and 0.11471 values between 0.11 and 0.23128 values between 0.23 and 0.3543 values between 0.35 and 0.4723 values between 0.47 and 0.597 values between 0.59 and 0.716 values between 0.71 and 0.835 values between 0.83 and 0.962 values between 0.96 and 1.081 values between 1.08 and 1.201 values between 1.32 and 1.441 values between 1.44 and 1.561 values between 1.56 and 1.682 values between 1.68 and 1.802 values between 1.80 and 1.921 values between 2.17 and 2.291 values between 3.01 and 3.13offnegativepositiveoften

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. sburg
  2. surrender
  3. voluntarily
  4. ve
  5. udi
  6. ม
  7. 誓
  8. volunteered
  9. obe
  10. VO

Suppresses

  1. ensen
  2. seau
  3. enet
  4. 眼界
  5. 发展格局
  6. educated
  7. 的眼光
  8. rium
  9. SCI
  10. 瘾