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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Initial Gemma-2B-IT SAEs
    3. Gemma-2B-IT
    4. Residual Stream
    5. 12-RES-JB
    6. 757
    Prev
    Next
    INDEX
    Explanations

    mentions of specific medical devices or procedures

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot

    information about financial and numerical figures, particularly dollar amounts and numerical measurements.

    oai_token-act-pair · claude-3-7-sonnet-20250219Triggered by @eekay

    proper nouns and terms related to patented inventions or intellectual property.

    oai_token-act-pair · deepseek-r1Triggered by @eekay

    The neuron responds to expository passages packed with quantitative details—numbers, measurements, dates, statistics, and other technical or factual specifications.

    oai_token-act-pair · o4-miniTriggered by @eekay
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    jbloom/Gemma-2b-IT-Residual-Stream-SAEs/gemma_2b_it_blocks.12.hook_resid_post_16384
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    HuggingFaceFW/fineweb
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.12.hook_resid_post
    Hook Layer
    12
    Architecture
    standard
    Context Size
    1,024
    Dataset
    Skylion007/openwebtext
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    1150
    +0.20
    0.8%
    599
    +0.19
    0.7%
    50
    +0.16
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    1499
    +0.20
    0.08
    1531
    +0.19
    0.04
    1948
    +0.16
    0.06
    Negative Logits
    <bos>
    -0.87
     سكانية
    -0.73
    GraphicsUnit
    -0.56
    MessageOf
    -0.55
    ⓧ
    -0.55
    RectangleBorder
    -0.55
    hyrchwyd
    -0.53
     PLWABN
    -0.52
    InputBorder
    -0.52
     saurait
    -0.49
    POSITIVE LOGITS
     unlaw
    0.65
     impractica
    0.61
     Moslem
    0.59
     vectorial
    0.57
     seclu
    0.57
     renunciation
    0.56
     vicissitudes
    0.55
     Freq
    0.53
     shenan
    0.53
     stratigraph
    0.53
    Activations Density 1.927%

    No Known Activations