Neuron Explorer

Qwen3-0.6B

"similar" after "are"

A neuron in layer 0 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 "similar" when it follows "are" or "is", as in "these habitats are similar to". It was strongly negative on the word "Devices" in phrases like "Fish Aggregating Devices". Turning it up gave text that did not clearly relate to the trigger.

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

    characterized viruses remain antigenically similar to the WHO recommended vaccines for the 2010-201

  2. #2 peak

    be fundamentally fair. Anything the police do that suggest which person in the lineup or photographic lineup is the suspect violates due process. There should be enough people who are similar in appearance in the lineup to force the witness to demonstrate his or her memory and

  3. #3 peak

    Conclusions Guidelines for AOM from developing and developed countries are similar in many aspects, with variation in specific recommendations, due to local epidemiology and

  4. #4 peak

    Irenaeus says:⏎"Those who are called Ebionites agree that the world was made by God; but their opinions with respect to the Lord are similar to those of Cerinthus and Carpocrates. They use the Gospel according to

  5. #5 peak

    , a family history of SAD, or even as simple as living in a state farther from the equator with less sunlight.⏎Symptoms of SAD are similar to depression, including anxiety, depressive feelings, appetite and weight changes, sleep changes

  6. #6 peak

    are - in grassy, open-canopied plant communities, or fire-adapted savanna-like habitats. These habitats are similar to the steppe and thorn scrub environments of western North America in which the

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: in the context of the past, the current, or the future. For example, if you have a problem that involves some complex, or some abstract, or some abstract, or some complex, or some complex, or some complex, or

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

    scientists from adequately interpreting the observed movements. Ideas for future research to studying tuna behaviour near FADs are discussed.⏎Key words: tuna / Fish Aggregating Devices / behaviour / horizontal movements / ultrasonic telemetry / French Polynesia⏎©

  2. #2 peak

    using ultrasonic transmitters in French Polynesia between 1985 and 1997. Nine fish were caught near Fish Aggregating Devices (FADs) while the other two were tracked in coastal areas without F

  3. #3 peak

    ⏎Advanced Exercise On English Tenses Mixed Online Exercise⏎Grammar And Vocabulary Worksheets For English⏎All Worksheets Are In Format And Can Be Viewed And Printed On All Devices Downloads Are Available Free Of Charge As Long As You Do Not Remove The Link To

  4. #4 peak

    of communication between two people in different locations. Letters are still used, particularly by businesses, for official notifications. This is because of three main advantages: No special device needed – almost everybody

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

None: this is the first layer.

Feeds later layers

  1. L17:127 weakly suppresses the word 'would' before a verb
  2. L19:2560 weakly excites
  3. L8:2789 weakly excites
  4. L3:1992 weakly excites
  5. L14:2425 weakly suppresses
  6. L11:1236 weakly suppresses
  7. L2:911 weakly suppresses
  8. L5:3039 weakly suppresses
  9. L1:2691 weakly excites
  10. L11:2545 weakly excites
  11. L8:769 weakly excites
  12. L8:660 weakly excites
  13. L23:2721 weakly excites
  14. L25:1631 weakly excites
  15. L26:752 weakly excites
  16. L7:788 weakly suppresses
  17. L25:728 weakly excites
  18. L26:55 weakly suppresses
  19. L26:912 weakly excites
  20. L20:112 weakly excites
  21. L15:663 weakly excites
  22. L25:930 weakly excites
  23. L6:2307 weakly suppresses
  24. L6:1756 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 -2.62 and -2.532 values between -2.24 and -2.141 values between -0.89 and -0.793 values between -0.79 and -0.706 values between -0.70 and -0.603 values between -0.60 and -0.502 values between -0.50 and -0.4114 values between -0.41 and -0.3125 values between -0.31 and -0.2186 values between -0.21 and -0.12753 values between -0.12 and -0.022743 values between -0.02 and 0.08222 values between 0.08 and 0.1731 values between 0.17 and 0.278 values between 0.27 and 0.372 values between 0.37 and 0.462 values between 0.46 and 0.561 values between 0.65 and 0.752 values between 0.75 and 0.851 values between 1.14 and 1.23offnegativepositiveoften

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. undry
  2. fullscreen
  3. ENT
  4. 哆
  5. ombres
  6. 哩
  7. ovable
  8. مة
  9. 到底
  10. ActivityResult

Suppresses

  1. caught
  2. rx
  3. ifix
  4. рег
  5. 牝
  6. liers
  7. urls
  8. ookie
  9. artic
  10. breaker