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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    EXPLANATION TYPE
    np_max-act
    Description
    A Neuronpedia original that forces concise explanations and shows the model the top activating tokens and texts. A simpler version of np_max-act-logits.
    Author
    Neuronpedia
    URL
    https://github.com/hijohnnylin/automated-interpretability/blob/917b11e38111c43526fe03ae6094a7081aeb982a/neuron_explainer/explanations/explainer.py#L1181
    Settings
    Activations shown = 24 tokens around max act. Shows model the max activating token too.
    Recent Explanations
    positive sentiment/praise
    gemini-2.0-flash
     the Bronx, spoke very highly of the sensational freshman’
    Neuronpedia logo
    GEMMA-2-2B
    12-GEMMASCOPE-RES-1M
    INDEX 506
    clever
    gemini-2.0-flash
     if need be. This clever solution pleased most players,
    Neuronpedia logo
    GEMMA-2-2B
    12-GEMMASCOPE-RES-1M
    INDEX 502
    prepositions
    gemini-2.0-flash
    ↵↵/*↵ * Called from a backedge in the
    Neuronpedia logo
    GEMMA-2-2B
    12-GEMMASCOPE-RES-1M
    INDEX 501
    Relationships and actions
    gemini-2.0-flash
    technical: it has to do with readability of your resultant
    Neuronpedia logo
    GEMMA-2-2B
    12-GEMMASCOPE-RES-1M
    INDEX 500
    +
    gemini-2.0-flash
    }=↵-(\gamma + i \Delta_e
    Neuronpedia logo
    GEMMA-2-2B
    2-RES-MATRYOSHKA-DC
    INDEX 1
    Words ending in "ing"
    gemini-2.0-flash
     yours.↵Since correct signing requires some sort of hashing
    Neuronpedia logo
    GEMMA-2-2B
    2-RES-MATRYOSHKA-DC
    INDEX 0
    code and numerical references
    gemini-2.0-flash
    .5281/
    Neuronpedia logo
    GEMMA-2-2B
    12-GEMMASCOPE-RES-1M
    INDEX 0
    prepositions
    gemini-2.0-flash
    Post navigation↵↵One thought on “Do We Remember:
    Neuronpedia logo
    GEMMA-2-2B
    15-GEMMASCOPE-TRANSCODER-16K
    INDEX 16316
    legal justification
    gemini-2.0-flash
     homicide was ruled to be justified in 236
    Neuronpedia logo
    GEMMA-2-2B
    15-GEMMASCOPE-TRANSCODER-16K
    INDEX 16342
    research/analysis
    gemini-2.0-flash
    power↵doctrine’s excesses, both
    Neuronpedia logo
    GEMMA-2-2B
    15-GEMMASCOPE-TRANSCODER-16K
    INDEX 16331
    CACHE
    gemini-2.0-flash
    ) {↵        return CACHE.get(id);
    Neuronpedia logo
    GEMMA-2-2B
    15-GEMMASCOPE-TRANSCODER-16K
    INDEX 16363
    a
    gemini-2.0-flash
    re)^{a}})_{a\in A})$ is
    Neuronpedia logo
    GEMMA-2-2B
    15-GEMMASCOPE-TRANSCODER-16K
    INDEX 16365
    providing information and guidance
    gemini-2.0-flash
     vivo and have the potential inform therapies for treatment of pathologies
    Neuronpedia logo
    GEMMA-2-2B
    15-GEMMASCOPE-TRANSCODER-16K
    INDEX 16347
    Latex/Math formulas
    gemini-2.0-flash
     \, \deg^L \, {[ \widetilde{
    Neuronpedia logo
    GEMMA-2-2B
    15-GEMMASCOPE-TRANSCODER-16K
    INDEX 16292
    choices and best options
    gemini-2.0-flash
     this quiz to find the best dress silhouette for your personality
    Neuronpedia logo
    GEMMA-2-2B
    15-GEMMASCOPE-TRANSCODER-16K
    INDEX 16375
    references to time
    gemini-2.0-flash
     growing up. But right now, she’s got
    Neuronpedia logo
    GEMMA-2-2B
    15-GEMMASCOPE-TRANSCODER-16K
    INDEX 16287
    Text excerpts/snippets
    gemini-2.0-flash
     two reliable alibis.↵↵Despite Pan's moderately
    Neuronpedia logo
    GEMMA-2-2B
    15-GEMMASCOPE-TRANSCODER-16K
    INDEX 16315
    relationships and community
    gemini-2.0-flash
     link and dontate your time, money and love to
    Neuronpedia logo
    GEMMA-2-2B
    15-GEMMASCOPE-TRANSCODER-16K
    INDEX 16326
    conflicts and organizations
    gemini-2.0-flash
    , which are totally convinced of their moral cause and use
    Neuronpedia logo
    GEMMA-2-2B
    15-GEMMASCOPE-TRANSCODER-16K
    INDEX 16348
    before
    gemini-2.0-flash
    \! ]$ denotes as before the set of vertices (
    Neuronpedia logo
    GEMMA-2-2B
    15-GEMMASCOPE-TRANSCODER-16K
    INDEX 16301