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. 34
    Prev
    Next
    INDEX
    Explanations

    words and phrases associated with conjunctions and connections between ideas

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

    Adjectives after conjunctions and commas

    np_acts-logits-general · gemini-2.0-flash

    descriptive adjectives in a list

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

    positive evaluative adjectives and praise words used to describe people or experiences.

    oai_token-act-pair · claude-4-5-haikuTriggered by @kparkhamchuk

    The conjunction "and" appearing between two adjectives, adverbs, or descriptive phrases that together characterize or modify a noun or concept. This pattern shows "and" functioning as a coordinating connector between parallel descriptive elements in a series, particularly in contexts emphasizing compound qualities or paired characteristics.

    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
    Autoritní
    -0.57
    FULLY
    -0.55
     subordination
    -0.53
    完全に
    -0.52
     apret
    -0.52
     esercit
    -0.52
    ably
    -0.52
     समीक्षाओं
    -0.51
     通販
    -0.51
    parsedMessage
    -0.51
    POSITIVE LOGITS
     efficient
    0.78
     effective
    0.76
    ')):
    0.71
    __":
    
    0.70
     productive
    0.69
    ']))
    
    0.69
     reliable
    0.67
    __':
    
    0.67
     decisive
    0.66
    ".
    
    0.66
    Activations Density 0.277%

    No Known Activations