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. 29366
    Prev
    Next
    INDEX
    Explanations

    references to historical events or concepts

    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
    +0.20
    1.1%
    100
    +0.14
    0.8%
    28
    +0.12
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    424
    +0.20
    0.01
    360
    +0.14
    0.02
    30
    +0.12
    0.02
    Negative Logits
    fficient
    -1.78
     thinner
    -1.66
     evenly
    -1.61
     treated
    -1.56
     skilled
    -1.54
     done
    -1.47
    wers
    -1.46
    akin
    -1.41
     tolerate
    -1.41
    vious
    -1.41
    POSITIVE LOGITS
    ¡
    1.75
    grounds
    1.74
    ités
    1.66
    isation
    1.66
    ignment
    1.59
    oire
    1.58
    icity
    1.55
     affairs
    1.51
     atroc
    1.46
    istically
    1.45
    Activations Density 0.201%

    No Known Activations