© 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
    2. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    3. Gemma-2-2B
    4. Residual Stream - 65k
    5. 20-GEMMASCOPE-RES-65K
    6. 16
    Prev
    Next
    INDEX
    Explanations

    words related to pistols and similar firearm terms

    oai_token-act-pair · gpt-4o-miniTriggered by @bot

    references to a specific plant species

    oai_token-act-pair · gpt-4o-miniTriggered by @bot

    pistol or piss-taking

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

    The highlighted tokens represent partial words or word fragments that appear within larger words. These fragments are typically 2-4 characters long and occur mid-word across diverse contexts including plant species names (Pistacia), proper nouns (Kishwaukee, Sisk), technical terms (pistol, memristance), and informal language (pis). The pattern suggests identification of morphologically interesting substrings or phonetic components embedded within words across different text domains.

    eleuther_acts_top20 · claude-4-5-haikuTriggered by @kparkhamchuk
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-2B @ 20-gemmascope-res-65k
    Configuration
    google/gemma-scope-2b-pt-res/layer_20/width_65k/average_l0_114
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    65,536
    Data Type
    float32
    Hook Name
    blocks.20.hook_resid_post
    Hook Layer
    20
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    WithIOException
    -1.22
    parsedMessage
    -1.01
     estimés
    -0.97
    oredCriteria
    -0.96
    دانشنامهٔ
    -0.94
    LookAnd
    -0.93
     disambiguazione
    -0.91
    InitVars
    -0.90
    ItemBackground
    -0.89
    endphp
    -0.87
    POSITIVE LOGITS
    iss
    0.62
    niss
    0.61
    ISC
    0.61
    Nis
    0.57
     ISS
    0.56
     Pitman
    0.55
    Bis
    0.54
    son
    0.53
     pis
    0.53
     CIS
    0.53
    Activations Density 0.077%

    No Known Activations