© 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
    2. Gemma-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 66121
    Prev
    Next
    INDEX
    Explanations

    messing up or screwing up

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

    The neuron detects colloquial multiword expressions that convey doing something badly or causing trouble—phrases like “messed up,” “screwed up,” “fucked with,” or “mess with.”

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     wou
    -1.16
     hilfreich
    -1.13
    ский
    -1.12
     actualiza
    -1.11
    ")==
    -1.10
    仅仅
    -1.09
     nėra
    -1.08
     updateUser
    -1.06
    只想
    -1.05
    ная
    -1.04
    POSITIVE LOGITS
    λογ
    1.29
    ActionCreators
    1.18
    },
    1.16
    してみた
    1.16
    ariate
    1.13
     $
    1.12
    雫
    1.10
    してみる
    1.08
    лада
    1.08
    âl
    1.08
    Activations Density 0.034%

    No Known Activations