Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    INDEX
    Explanations

    would/could/might/should

    np_acts-logits-general · gemini-2.5-flash-lite

    discourse markers that signal explanation, contrast, or hypotheticals, along with first‑person metacommentary in analytical or expository text.

    oai_token-act-pair · gpt-5Triggered by @yooniel31
    New Auto-Interp
    Negative Logits
     jangan
    0.45
     Jangan
    0.45
     Sometimes
    0.44
     ہے
    0.44
     når
    0.44
     není
    0.44
     Když
    0.43
     Когда
    0.43
     পাইয়া
    0.43
     Når
    0.43
    POSITIVE LOGITS
     would
    3.20
    would
    3.13
     Would
    2.86
    Would
    2.84
     WOULD
    2.75
     sarebbero
    2.55
     seraient
    2.47
     serait
    2.39
     sarebbe
    2.38
     würde
    2.31
    Act Density 0.087%

    No Known Activations