Neuron Explorer

Qwen3-0.6B

'yes' answering a question

A neuron in layer 17 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the word 'Yes' (or the comma right after it) when it answers a yes or no question. Turning it down made the model keep repeating 'is it true' and 'the statement is true', which fits a theme about truth and yes/no answers.

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

    children yourself.⏎1. What is prayer? (Prayer is spiritual communication between ourselves and God.)⏎2. Can you pray when you are in trouble? (Yes.)⏎3. Can you pray when you are happy? (Yes.)⏎4.

  2. #2 peak

    (at no extra cost to you). As an Amazon Associate I earn from qualifying purchases. Please read my disclaimer for additional details.⏎Can goats eat strawberries?⏎Yes, goats can eat strawberries, however, not all goats are fond of their sour and

  3. #3 peak

    using teeth to process their meals, they also use similar movements in order to communicate with each other. How amazing is that?⏎Do Koi Lose Their Teeth?⏎Yes, koi fish may lose their teeth every once in a while. But that

  4. #4 peak

    . This search is not the biggest one, but the look of it is great.⏎But you are referring to a large diameter compared to the depth of it.⏎Yes, there is a limit. The image on the left is a diagram of an

  5. #5 peak

    rowing, canoeing and kayaking?⏎Is there, actually? Is rowing, canoeing and kayaking very different from each other? The answer is yes, they are. These are not three different techniques of operating the same boat.

  6. #6 peak

    your left pinky finger… where the nail meets the skin…” Do you think of it? Did you think of your pinky because I mentioned it? You did that because that is how quickly our brain responds.⏎Tip 1. Keep a

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: is it true? Is it not? The question is not the question of the question of the question? Is it not the question? Is it not the first of the first? I am not the first? Is it not the first?

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

    bloom booster, at most, only contains only trace elements of essential micro and macronutrients.⏎Plant Magic is an example of a base nutrient plant fertiliser.⏎To be a PK booster all it has to do is is contain phosphorus and

  2. #2 peak

    ``. Per urllib2, it has an ``info()`` method⏎containing things like headers, etc.⏎Upon success, if you set ``parse_xml_document=True``, this will⏎contain the parsed xml_document. I'll take care of

  3. #3 peak

    slope of the diagram.⏎A scatter diagram is composed of a horizontal axis containing the measured values of one variable and a vertical axis representing the measurements of the other variable.⏎The purpose of the scatter diagram is to display what happens to one variables when another

  4. #4 peak

    ), foreign languages, health and physical education<|endoftext|>'Gay' is the word used to describe the physical and emotional attraction of a person to people of the same sex. It is often used to describe men or boys who are sexually attracted to other males

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. L15:2721 strongly excites
  2. L16:1186 suppresses
  3. L16:1806 suppresses
  4. L10:263 suppresses
  5. L13:2939 excites
  6. L12:1752 excites
  7. L15:1667 suppresses
  8. L7:1670 suppresses
  9. L11:1592 weakly suppresses
  10. L15:2169 weakly suppresses
  11. L12:237 weakly suppresses
  12. L15:2385 weakly excites
  13. L11:1678 weakly suppresses
  14. L13:2761 weakly suppresses
  15. L7:512 weakly excites
  16. L3:557 weakly suppresses
  17. L13:438 weakly suppresses
  18. L10:1500 weakly suppresses
  19. L9:1683 weakly excites
  20. L16:1671 weakly excites
  21. L15:2412 weakly suppresses
  22. L16:2793 weakly suppresses
  23. L16:918 weakly suppresses
  24. L15:980 weakly excites

Feeds later layers

  1. L18:1784 strongly suppresses
  2. L23:968 strongly suppresses
  3. L18:1531 strongly suppresses
  4. L21:1538 strongly suppresses
  5. L21:1341 suppresses
  6. L25:636 suppresses
  7. L23:1463 suppresses
  8. L27:1796 suppresses
  9. L20:782 suppresses
  10. L21:1132 suppresses
  11. L22:2619 suppresses
  12. L20:1263 suppresses
  13. L18:2484 suppresses
  14. L18:1582 suppresses
  15. L18:2036 suppresses
  16. L24:1255 suppresses
  17. L21:1592 suppresses
  18. L21:2340 suppresses
  19. L21:1044 suppresses
  20. L26:1185 suppresses
  21. L22:1933 suppresses
  22. L20:1479 suppresses
  23. L19:2544 suppresses
  24. L18:1178 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 -7.83 and -7.571 values between -7.32 and -7.061 values between -6.30 and -6.041 values between -6.04 and -5.791 values between -5.79 and -5.531 values between -5.53 and -5.281 values between -5.02 and -4.772 values between -3.75 and -3.492 values between -3.24 and -2.981 values between -2.98 and -2.731 values between -2.22 and -1.963 values between -1.96 and -1.712 values between -1.71 and -1.456 values between -1.45 and -1.206 values between -1.20 and -0.944 values between -0.94 and -0.697 values between -0.69 and -0.4352 values between -0.43 and -0.181044 values between -0.18 and 0.081912 values between 0.08 and 0.33689 values between 0.33 and 0.59144 values between 0.59 and 0.8422 values between 0.84 and 1.102 values between 1.10 and 1.351 values between 1.35 and 1.611 values between 2.12 and 2.37offnegativepositiveoften

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. icut
  2. 让我们
  3. blurry
  4. 模糊
  5. iculo
  6. iele
  7. �示
  8. 大人
  9. ikan
  10. 多重

Suppresses

  1. True
  2. true
  3. truth
  4. True
  5. yes
  6. (True
  7. TRUE
  8. true
  9. =True
  10. False