© 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. 9979
    Prev
    Next
    INDEX
    Explanations

    future tense verbs and phrases related to expectations or requirements

    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
    219
    +0.13
    0.7%
    499
    +0.10
    0.6%
    108
    +0.10
    0.6%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    219
    +0.13
    0.10
    388
    +0.10
    0.09
    317
    +0.10
    0.08
    Negative Logits
    ¼
    -2.14
    ictionaries
    -1.76
    ĩ
    -1.59
    Ļª
    -1.57
    İ
    -1.56
     followed
    -1.52
    enza
    -1.50
    #,
    -1.47
    ľĵ
    -1.47
    ľ
    -1.46
    POSITIVE LOGITS
     need
    1.58
     observer
    1.49
     expect
    1.46
     wonder
    1.45
     forum
    1.40
    stant
    1.35
    want
    1.33
     insult
    1.32
     encounter
    1.32
     target
    1.32
    Activations Density 1.263%

    No Known Activations