© 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. 2-RES-SM
    11. 28366
    Prev
    Next
    INDEX
    Explanations

    the term "like" as a comparison or preference indicator

    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
    47
    +0.12
    0.7%
    81
    +0.11
    0.6%
    120
    +0.11
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    407
    +0.12
    0.02
    274
    +0.11
    0.03
    91
    +0.11
    0.02
    Negative Logits
    ĺ
    -1.86
    ¤
    -1.86
    Ľ
    -1.80
    ĵ
    -1.78
    Ł
    -1.70
    IJ
    -1.70
    athe
    -1.67
    ette
    -1.64
    ¨
    -1.60
    «
    -1.58
    POSITIVE LOGITS
     http
    1.72
    :"
    1.70
     https
    1.68
    ://
    1.65
    :'
    1.60
    liament
    1.52
     |\
    1.51
    urger
    1.51
     "^
    1.49
    rror
    1.49
    Activations Density 0.084%

    No Known Activations