Neuron Explorer

Qwen3-0.6B

'out' in 'X out there' phrases

A neuron in layer 24 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 'out' in phrases like 'many calculators out there' or 'plenty of types of support out there'. When it was turned up, the model just kept writing 'out out out'.

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

    a spider use such effective hunting strategies to hunt a plant."<|endoftext|>You want to live a healthier life, but where do you start? There are so many people out there telling you what you should and shouldn’t do, what you should and shouldn

  2. #2 peak

    we don’t control very much. TEF and TEA we can make changes to. There are many calculators out there which can help you determine your energy expenditure. One that seems<|endoftext|>Italian scientists

  3. #3 peak

    are to<|endoftext|>It takes a while before one can master using a lathe. The following guide is for a wood working lathe, but there are many types out there.⏎First off, you should don safety glasses. Remove any jewelry prior to

  4. #4 peak

    an opportunity to share positive experiences⏎- Provide emotional support and allow you to support others⏎- There are plenty of different types of support out there, and a Mental Health First Aider can help you access them.⏎-

  5. #5 peak

    your employees⏎Do a small search online, and you’ll soon find out that there are many different technical skills out there. But which ones are the most important for today’s employees? Here are

  6. #6 peak

    the U.S. business forecasting firm DRI-WEFA. “A lot of prominent geologists just laugh at this. There are wolves out there, but if you keep crying wolf and no wolves show up, you start

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 first and first really really first. I am really, really, really first in... what? I just really just really first. I am first in the first of all first, first of all... I am really first in... the

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

    you join two threaded rods together?⏎Couplings nut are used to join two threaded rods<|endoftext|>Expanding the functional human mitochondrial DNA database by the establishment of primate xenomitochondrial cybrids.⏎The nuclear and mitochondrial genomes co

  2. #2 peak

    preen its head feathers. Usually preening is done with the bill. See HERE in the case of the Red-breasted Parakeet (Psittacula alexandri). On the other hand, with aboreal birds, they scratch with

  3. #3 peak

    Parks System are in Eureka. As entrepôt for hundreds of lumber mills that once existed in the area, the city played a leading role in the historic West Coast lumber trade. The entire city is a state historic landmark, which has hundreds

  4. #4 peak

    , and gorilla replaced the human mtDNA and restored respiration to essentially normal levels. Mitochondrial protein synthesis was also unaltered in successful "xenomitochondrial cybrids." The abrupt failure of mtDNA from primate species

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. L19:1109 suppresses
  2. L8:1288 weakly suppresses
  3. L22:2450 weakly suppresses
  4. L3:803 weakly excites
  5. L23:2341 weakly excites
  6. L14:2471 weakly suppresses
  7. L12:1320 weakly excites
  8. L20:1230 weakly excites
  9. L20:1051 weakly suppresses
  10. L23:476 weakly suppresses
  11. L3:1418 weakly suppresses
  12. L7:795 weakly suppresses
  13. L7:2644 weakly excites
  14. L20:669 weakly suppresses
  15. L16:2900 weakly suppresses
  16. L10:1908 weakly suppresses 'for' after study resource nouns
  17. L23:3038 weakly suppresses
  18. L11:1581 weakly suppresses
  19. L20:1402 weakly excites
  20. L7:2582 weakly excites
  21. L21:905 weakly suppresses
  22. L4:1588 weakly suppresses
  23. L3:2413 weakly suppresses
  24. L19:1450 weakly excites

Feeds later layers

  1. L27:850 strongly excites
  2. L26:1474 excites
  3. L27:1301 excites
  4. L25:1453 weakly suppresses
  5. L27:1008 weakly excites
  6. L25:1701 weakly suppresses
  7. L25:2642 weakly suppresses
  8. L25:666 weakly excites
  9. L27:2711 weakly suppresses
  10. L26:2580 weakly excites
  11. L25:774 weakly excites
  12. L27:1428 weakly suppresses
  13. L27:1251 weakly excites
  14. L25:2608 weakly excites
  15. L26:1987 weakly suppresses
  16. L25:1681 weakly suppresses
  17. L25:2257 weakly suppresses
  18. L26:2377 weakly suppresses
  19. L27:1728 weakly suppresses
  20. L26:1521 weakly excites
  21. L25:689 weakly suppresses
  22. L25:3040 weakly excites
  23. L27:845 weakly suppresses
  24. L25:869 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 -21.47 and -20.251 values between -14.16 and -12.942 values between -12.94 and -11.731 values between -10.51 and -9.294 values between -9.29 and -8.076 values between -8.07 and -6.853 values between -6.85 and -5.647 values between -5.64 and -4.4213 values between -4.42 and -3.2030 values between -3.20 and -1.98102 values between -1.98 and -0.763206 values between -0.76 and 0.46423 values between 0.46 and 1.6758 values between 1.67 and 2.8920 values between 2.89 and 4.1111 values between 4.11 and 5.336 values between 5.33 and 6.554 values between 6.55 and 7.764 values between 7.76 and 8.982 values between 8.98 and 10.201 values between 10.20 and 11.421 values between 11.42 and 12.631 values between 21.16 and 22.381 values between 26.03 and 27.25offnegativepositiveoften

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. out
  2. (out
  3. -out
  4. out
  5. 出
  6. →out
  7. *out
  8. Out
  9. =out
  10. Out

Suppresses

  1. ron
  2. олж
  3. ihu
  4. RON
  5. КО
  6. orr
  7. apesh
  8. iah
  9. Gebä
  10. amin