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. 40
    Prev
    Next
    INDEX
    Explanations

    phrases related to hybrid combinations of concepts or elements in various contexts

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

    combination

    np_max-act-logits · gpt-4o-miniTriggered by @johnny

    Combination of multiple things

    np_acts-logits-general · gemini-2.0-flash

    combination or combinations

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

    The marked tokens represent short phrasal or conjunctive elements that connect ideas or provide alternatives within sentences. These include coordinating conjunctions (or, and), prepositions (with, by), and short phrases that express combination, modification, or alternative possibilities (a hybrid, a synergistic, a linear, their cooperation). They function as cohesive bridges between clauses or noun phrases, often indicating multiple options, concurrent actions, or compositional relationships in technical, descriptive, and expository writing.

    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
     '\\;'
    -0.68
    AndEndTag
    -0.60
     newOwner
    -0.58
    AsUp
    -0.57
    elemField
    -0.55
    reactstrap
    -0.54
     disambiguazione
    -0.53
    UnsafeEnabled
    -0.53
     stanov
    -0.51
    InjectAttribute
    -0.51
    POSITIVE LOGITS
     combination
    3.32
    combination
    3.03
     Combination
    3.00
    Combination
    2.93
     combinations
    2.86
     COMBIN
    2.83
     combining
    2.75
     Kombination
    2.67
     combine
    2.67
     combinatie
    2.66
    Activations Density 1.023%

    No Known Activations