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. 29410
    Prev
    Next
    INDEX
    Explanations

    elements related to programming and technical specifications

    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
    419
    +0.18
    1.0%
    330
    +0.14
    0.8%
    302
    +0.14
    0.8%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    330
    +0.18
    0.13
    222
    +0.14
    0.08
    431
    +0.14
    0.06
    Negative Logits
    Developer
    -1.61
     respectively
    -1.50
    TestCase
    -1.43
    Additional
    -1.41
    pent
    -1.40
    qualified
    -1.38
    Person
    -1.35
     Petersburg
    -1.33
    Solution
    -1.32
    Figure
    -1.32
    POSITIVE LOGITS
    ncia
    1.60
     loud
    1.49
     noise
    1.45
    =”
    1.43
     coincidence
    1.42
     autoc
    1.35
     attachments
    1.34
     held
    1.33
     asymmetry
    1.32
    nae
    1.32
    Activations Density 5.256%

    No Known Activations