© 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-3-12B
    3. 24-GEMMASCOPE-2-RES-16K
    4. 3991
    Prev
    Next
    INDEX
    Explanations

    broken down

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

    The neuron strongly activates on words that introduce or carry key relational attributions in a sentence—especially relative‐clause markers (“who,” “which”) and the verbs/adjectives that follow them (e.g. “described,” “responsible,” “owns,” “projected”).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-12b-pt/resid_post/layer_24_width_16k_l0_medium
    Prompts (Dashboard)
    392,802 prompts, 256 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    });
    0.64
    TimeTo
    0.58
    ได้
    0.56
    ennia
    0.55
    inities
    0.53
    }');
    0.53
     ความ
    0.52
    的原因
    0.52
     ダイニング
    0.52
    iness
    0.52
    POSITIVE LOGITS
     notoriously
    0.97
     cuja
    0.96
     которого
    0.94
     którego
    0.94
     cujo
    0.92
     whose
    0.91
     normalerweise
    0.90
     lequel
    0.90
     cuya
    0.89
     kjent
    0.89
    Activations Density 1.073%

    No Known Activations