© 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. 12317
    Prev
    Next
    INDEX
    Explanations

    elements indicating emotional tension and physical conflict

    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
    56
    +0.12
    0.7%
    427
    +0.11
    0.6%
    306
    +0.11
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    7
    +0.12
    0.03
    164
    +0.11
    0.12
    163
    +0.11
    0.09
    Negative Logits
     whom
    -2.05
     behalf
    -2.00
    $.[]{
    -1.83
    )\|_{
    -1.72
    selves
    -1.55
     who
    -1.54
     instance
    -1.54
    +.
    -1.50
    .\
    -1.50
    soever
    -1.50
    POSITIVE LOGITS
     began
    2.14
     faded
    2.12
     froze
    2.08
     slowed
    2.08
     erupted
    2.05
     stops
    2.04
     seemed
    2.04
     came
    1.99
     starts
    1.97
     grew
    1.97
    Activations Density 1.486%

    No Known Activations