© 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. Under Peer Review · Sparse Autoencoders for Pythia-70M-Deduped
    5. Pythia-70M-Deduped
    7. Residuals
    9. 4-RES-SM
    11. 11648
    Prev
    Next
    INDEX
    Explanations

    references to language-related concepts and vocabulary influences

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    ctigges/pythia-70m-deduped__res-sm_processed/4-res-sm
    Prompts (Dashboard)
    32,768 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    32,768
    Data Type
    torch.float32
    Hook Name
    blocks.4.hook_resid_post
    Hook Layer
    4
    Architecture
    standard
    Context Size
    128
    Dataset
    EleutherAI/the_pile_deduplicated
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    274
    +0.16
    0.9%
    40
    +0.13
    0.7%
    365
    +0.12
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    379
    +0.16
    0.07
    135
    +0.13
    0.07
    58
    +0.12
    0.08
    Negative Logits
     apoptotic
    -1.79
     apopt
    -1.75
     doxorubicin
    -1.74
     anode
    -1.73
     fiduc
    -1.67
    orbit
    -1.66
     Motors
    -1.65
    athione
    -1.64
    ¿½
    -1.61
     pic
    -1.57
    POSITIVE LOGITS
     languages
    4.36
     lingu
    3.95
     Language
    3.76
     language
    3.76
     dialect
    3.76
     English
    3.65
     translator
    3.65
     linguistic
    3.64
     grammar
    3.64
     translations
    3.39
    Activations Density 1.489%

    No Known Activations