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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Under Peer Review · Sparse Autoencoders for Pythia-70M-Deduped
    3. Pythia-70M-Deduped
    4. Residuals
    5. 2-RES-SM
    6. 11613
    Prev
    Next
    INDEX
    Explanations

    "up" and its associated variations or a sense of increase

    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
    317
    +0.15
    0.9%
    228
    +0.13
    0.7%
    162
    +0.11
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    41
    +0.15
    0.03
    467
    +0.13
    0.02
    66
    +0.11
    0.03
    Negative Logits
    ĥ½
    -2.57
    Ī
    -2.12
    ©
    -1.86
    §
    -1.86
    ¯
    -1.79
    Ń
    -1.78
    ¥
    -1.76
    ĸ´
    -1.65
    ĵ
    -1.65
    ī
    -1.63
    POSITIVE LOGITS
    dates
    2.92
    grades
    2.45
    dating
    2.12
    grade
    1.95
    ublic
    1.91
    graded
    1.83
    hill
    1.72
    stairs
    1.72
    ercase
    1.67
    grad
    1.67
    Activations Density 0.154%

    No Known Activations