© 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
    3. Joseph Bloom · Initial Gemma-2B-IT SAEs
    5. Gemma-2B-IT
    7. Residual Stream
    9. 12-RES-JB
    11. 13436
    Prev
    Next
    INDEX
    Explanations

    phrases containing the word "which" followed by various contextual information

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    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
    513
    +0.13
    0.4%
    871
    +0.10
    0.3%
    795
    +0.10
    0.3%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    513
    +0.13
    0.05
    795
    +0.10
    0.04
    528
    +0.10
    0.04
    Negative Logits
     kram
    -1.36
     lele
    -1.32
     hina
    -1.15
     gend
    -1.15
     meis
    -1.13
     ananas
    -1.11
     krab
    -1.10
     tomat
    -1.10
     karton
    -1.08
     saar
    -1.07
    POSITIVE LOGITS
     which
    0.82
     thereupon
    0.80
     tolerably
    0.77
     vainly
    0.77
     Which
    0.77
    Which
    0.73
     whither
    0.73
     apprehen
    0.71
     exasper
    0.70
     makes
    0.69
    Activations Density 0.134%

    No Known Activations