Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Under Peer Review · Sparse Autoencoders for Pythia-70M-Deduped
    3. Pythia-70M-Deduped
    4. Residuals
    5. 2-RES-SM
    6. 29172
    Prev
    Next
    INDEX
    Explanations

    compound terms and coding constructs from programming and data structures

    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
    271
    +0.22
    1.3%
    186
    +0.22
    1.3%
    235
    +0.17
    1.0%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    271
    +0.22
    -0.01
    231
    +0.22
    0.13
    210
    +0.17
    0.04
    Negative Logits
     Examination
    -1.62
     accompanies
    -1.50
     CIRCUIT
    -1.48
     exceeded
    -1.38
    ings
    -1.38
     resulted
    -1.34
     satisfied
    -1.34
    heta
    -1.31
    ]."
    -1.31
     complied
    -1.31
    POSITIVE LOGITS
    ĻĤ
    2.50
    ¶
    2.21
    ·¸
    2.13
    ³
    2.13
    Ĵ
    2.11
    ŀ
    2.10
    ļ
    2.10
    Ļª
    2.09
    ĭ
    2.04
    ľ
    2.04
    Activations Density 3.318%

    No Known Activations