© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW 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. 32153
    Prev
    Next
    INDEX
    Explanations

    sentences that convey important statements or conclusions

    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
    258
    +0.19
    1.1%
    203
    +0.12
    0.7%
    425
    +0.10
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    258
    +0.19
    0.07
    432
    +0.12
    0.04
    425
    +0.10
    0.06
    Negative Logits
    iNdEx
    -2.06
    ĻĤ
    -1.96
    breviations
    -1.78
    blockList
    -1.75
    istor
    -1.73
    (<
    -1.71
    jsfiddle
    -1.63
    optic
    -1.58
    istors
    -1.56
    goog
    -1.56
    POSITIVE LOGITS
     Pence
    1.53
     pains
    1.43
     breadth
    1.41
     distance
    1.38
    ÂĶ
    1.37
     Ocean
    1.36
     delight
    1.34
     Eve
    1.33
     anybody
    1.32
    sworth
    1.32
    Activations Density 0.346%

    No Known Activations