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

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-3-4B-IT
    5. 22-GEMMASCOPE-2-RES-16K
    7. 2497
    Prev
    Next
    INDEX
    Explanations

    The neuron flags strong profanity—especially multi‐word or intensified swears (e.g. “God damn,” “fuck,” “cunt”)—marking when highly offensive curse phrases occur.

    oai_token-act-pair · o4-miniTriggered by @drejcpesjakpesjak

    discussions of profanity and offensive speech, including meta-talk about speaking style and advice or warnings around using such language.

    oai_token-act-pair · gpt-5Triggered by @drejcpesjakpesjak
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-4b-it/resid_post/layer_22_width_16k_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
     운영
    0.89
    🔬
    0.87
    📈
    0.84
    🧪
    0.83
     bioinformatics
    0.82
     DeFi
    0.78
    ipynb
    0.74
     பணிகள்
    0.74
     OpenGL
    0.73
    📊
    0.73
    POSITIVE LOGITS
     utterances
    2.67
     utterance
    2.61
     phrases
    2.54
     frases
    2.33
     verbal
    2.30
     phrasing
    2.29
     speech
    2.27
     uttered
    2.26
     phrase
    2.18
     words
    2.15
    Activations Density 2.924%

    No Known Activations