© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    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. 2-RES-SM
    11. 30489
    Prev
    Next
    INDEX
    Explanations

    terms related to cyberattacks and security vulnerabilities

    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/2-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.2.hook_resid_post
    Hook Layer
    2
    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
    156
    +0.21
    1.2%
    186
    +0.20
    1.2%
    351
    +0.15
    0.9%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    186
    +0.21
    0.09
    88
    +0.20
    0.10
    480
    +0.15
    0.09
    Negative Logits
    £
    -3.35
    ij
    -3.25
    IJ
    -3.21
    ĸ´
    -3.14
    ĥ
    -3.07
    ŀ
    -2.96
    ļ
    -2.96
    ©
    -2.94
    ¨
    -2.93
    ¢
    -2.91
    POSITIVE LOGITS
     malware
    2.42
     malicious
    2.32
     fake
    2.14
     fraudulent
    2.13
     vulnerabilities
    2.07
     violation
    1.95
     misuse
    1.94
     breaches
    1.93
     fraud
    1.92
     conceal
    1.91
    Activations Density 5.302%

    No Known Activations