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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    3. Gemma-2-2B
    4. Residual Stream - 16k
    5. 20-GEMMASCOPE-RES-16K
    6. 39
    Prev
    Next
    INDEX
    Explanations

    details about product features and specifications

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

    Tokens before/after specific keywords

    np_acts-logits-general · gemini-2.0-flash

    User bank, program, absent

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

    Small function words and punctuation marks are marked as important across product descriptions, technical specifications, and informational texts. These include articles (the, a), prepositions (of, in, for), conjunctions (and, or), commas, and parentheses that structure lists and clarifications. The pattern suggests these tokens are significant for parsing compound noun phrases, feature enumerations, and technical specifications where grammatical connectors organize product attributes and comparisons.

    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-16k
    Configuration
    google/gemma-scope-2b-pt-res/layer_20/width_16k/average_l0_71
    Prompts (Dashboard)
    36,864 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    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
    ⓧ
    -0.71
     estimés
    -0.60
    obox
    -0.58
    IntoConstraints
    -0.58
     оригіналу
    -0.56
     EconPapers
    -0.56
    AccessorTable
    -0.54
     Hano
    -0.53
    Geplaatst
    -0.52
     ویکی‌پدی
    -0.51
    POSITIVE LOGITS
     programmed
    0.63
     functions
    0.62
     LCD
    0.61
     reset
    0.61
     Chwiliwch
    0.61
     memory
    0.59
     temperature
    0.59
     backlit
    0.59
    program
    0.59
     Program
    0.58
    Activations Density 0.237%

    No Known Activations