Neuron Explorer

Qwen3-0.6B

'any' after negatives or 'more'

A neuron in layer 17 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

The neuron lights up on the word 'any' in sentences with a 'not', 'no more', or 'any more' feel, such as 'do not suffer from any forms' or 'spend any more'. Turning it up did not produce a clear theme; the outputs became repetitive and shifted to other topics.

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

    out crimes of a sexual nature is held accountable. As stated on the Rape Crisis website, research has shown us that the majority of these individuals do not suffer from any forms of mental illness nor do they need psychiatric attention, they do not carry out

  2. #2 peak

    wasn't one today."⏎It is a question that will resonate with millions living in quarantine today – as people wake up every morning wondering if the day ahead will be any different from the 24 hours they have just endured.⏎But I have a

  3. #3 peak

    to then decide how to spend. Once your child runs out of cash in an envelope, they have met their budget for that pay period and are unable to spend any more in that category until the next pay period. Creating a challenge encourages your kids

  4. #4 peak

    30 research studies over the past decade that have looked at this issue and most have found that:⏎- People with brain tumors do not report using cell phones any more than people who don’t have tumors⏎- People

  5. #5 peak

    be able to send astronauts in another 50 years from now.⏎One thing’s for sure though – the fascination with space exploration isn’t likely to go away any time soon.⏎“There’s a driving desire within us to find out who we are

  6. #6 peak

    it. But like other multimodals, I can also learning mainly by reading or hearing. In other words, we are not restricted to our favorite learning styles any more that I'm restricted to my preferred flavor of ice cream (vanilla).⏎

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 have not been doing all the same as I used to. I am going to be doing all the same as I used before. I am going to be doing all the same as I used before. I I am going to be

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

    SA system in Australia's Hunter Valley by mid-2012 before going into full production.<|endoftext|>As the world increasingly looks to move beyond carbon fuel, some environmentalists view nuclear power as a key component of a cleaner fuel future. Others

  2. #2 peak

    gas that seeps into some homes, from the ground, through sump pumps, cracks in the foundation, etc., it is considered a health hazard.⏎Any of the beams that slope from the ridge of a roof to the eaves

  3. #3 peak

    Some of the rise in CO<|endoftext|>Therapeutic Resources welcomes therapists eligible for sponsorship to work

  4. #4 peak

    little city of Trier is great. It is a big honor if people pay attention to Trier.”⏎The statue has reportedly stirred up controversy among locals, with some citizens questioning why it is being created by a Chinese artist. Others worry that the

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. L16:960 excites the word "some" mid-sentence
  2. L3:914 excites "any" in negated sentences
  3. L16:342 excites
  4. L0:3014 weakly excites
  5. L15:280 weakly excites
  6. L11:267 weakly excites
  7. L15:957 weakly suppresses
  8. L8:126 weakly excites
  9. L8:10 weakly excites
  10. L8:7 weakly excites
  11. L16:282 weakly excites
  12. L16:1513 weakly suppresses opening parenthesis before a numbered list
  13. L9:59 weakly suppresses "a" meaning per (a day, a year)
  14. L14:2007 weakly suppresses
  15. L2:1985 weakly excites
  16. L8:14 weakly excites
  17. L4:402 weakly suppresses
  18. L0:1111 weakly suppresses
  19. L11:768 weakly suppresses
  20. L7:2075 weakly excites
  21. L15:122 weakly suppresses
  22. L2:493 weakly suppresses
  23. L4:1076 weakly excites
  24. L14:689 weakly suppresses

Feeds later layers

  1. L19:1180 suppresses
  2. L19:1023 excites
  3. L23:2339 excites
  4. L21:592 excites
  5. L25:1586 suppresses
  6. L18:216 suppresses
  7. L27:797 suppresses
  8. L20:889 suppresses
  9. L22:404 excites
  10. L21:1705 excites
  11. L18:314 weakly suppresses
  12. L20:835 weakly excites
  13. L19:1016 weakly suppresses
  14. L24:153 weakly suppresses
  15. L27:814 weakly excites
  16. L21:383 weakly excites
  17. L18:635 weakly excites
  18. L19:1539 weakly suppresses
  19. L25:390 weakly excites
  20. L23:79 weakly suppresses
  21. L26:628 weakly suppresses
  22. L26:1652 weakly excites
  23. L27:588 weakly excites
  24. L20:424 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 -2.91 and -2.0115 values between -2.01 and -1.101660 values between -1.10 and -0.192183 values between -0.19 and 0.7217 values between 0.72 and 1.638 values between 1.63 and 2.546 values between 2.54 and 3.452 values between 3.45 and 4.365 values between 4.36 and 5.274 values between 5.27 and 6.172 values between 6.17 and 7.082 values between 7.08 and 7.991 values between 11.63 and 12.541 values between 18.90 and 19.811 values between 32.53 and 33.44offnegativepositiveoften

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. place
  2. THING
  3. thing
  4. CAP
  5. one
  6. elines
  7. ifference
  8. lec
  9. uber
  10. thin

Suppresses

  1. astery
  2. -bodied
  3. appell
  4. himself
  5. an
  6. яем
  7. ас
  8. (days
  9. some
  10. -days