© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    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. 28897
    Prev
    Next
    INDEX
    Explanations

    instances of comparisons or contrasts in complex language

    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
    240
    +0.27
    1.6%
    156
    +0.22
    1.3%
    369
    +0.21
    1.2%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    240
    +0.27
    0.03
    369
    +0.22
    0.03
    125
    +0.21
    0.04
    Negative Logits
    teenth
    -1.49
     Fig
    -1.48
    documentclass
    -1.34
    ^âĪĴ
    -1.32
     0000000000000000000000000000000000
    -1.29
    nolimits
    -1.28
     trib
    -1.27
     amusement
    -1.27
    urane
    -1.26
    ftime
    -1.26
    POSITIVE LOGITS
    ²
    2.64
    Ĵ
    2.58
    »
    2.52
    Ĭ
    2.52
    º
    2.52
    ¢
    2.52
    Ľ
    2.51
    ĸ´
    2.50
    ¸
    2.49
    Ĺ
    2.49
    Activations Density 1.430%

    No Known Activations