© 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 EvalsExportsinterp-engineNEWAPI 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. 13545
    Prev
    Next
    INDEX
    Explanations

    expressions of the speaker's feelings and states of being

    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
    307
    +0.14
    0.8%
    214
    +0.13
    0.7%
    337
    +0.12
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    337
    +0.14
    0.10
    214
    +0.13
    0.09
    288
    +0.12
    0.10
    Negative Logits
    º
    -4.24
    ª
    -4.24
    ł
    -4.22
    »
    -4.03
    ¯
    -4.01
    Ĥ¬
    -4.00
    ¢
    -3.90
    »¿
    -3.87
    ĨĴ
    -3.83
      ↵
    -3.82
    POSITIVE LOGITS
     glad
    2.92
     sorry
    2.89
     proud
    2.72
     grateful
    2.64
     thankful
    2.58
     afraid
    2.52
     pleased
    2.47
     happy
    2.41
     aware
    2.41
     fortunate
    2.38
    Activations Density 0.470%

    No Known Activations