© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-3-27B-IT
    5. 40-GEMMASCOPE-2-RES-262K
    7. 1297
    Prev
    Next
    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
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-27b-it/resid_post/layer_40_width_262k_l0_medium
    Prompts (Dashboard)
    238,145 prompts, 512 tokens each
    Dataset (Dashboard)
    lmsys + oasst1
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    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
    Activations Density 0.087%

    No Known Activations