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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Open Source Sparse Autoencoders for all Residual Stream Layers of GPT2-Small
    3. GPT2-Small
    4. Residual Stream
    5. 4-RES-JB
    6. 21285
    Prev
    Next
    INDEX
    Explanations

    the phrase "die hard" used in various contexts, potentially related to strong beliefs or intensities

    oai_token-act-pair · gpt-3.5-turbo

    phrases related to mortality or dying

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 4-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.hook_resid_pre
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    Data Type
    torch.float32
    Hook Point
    blocks.4.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    4
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    eele
    -1.02
    oyer
    -0.85
    kefeller
    -0.82
    arthy
    -0.78
    yip
    -0.75
    orah
    -0.68
    android
    -0.67
    Cu
    -0.64
    soDeliveryDate
    -0.61
    ghai
    -0.60
    POSITIVE LOGITS
     ..........
    0.69
    iably
    0.67
     grace
    0.66
     inertia
    0.66
     reborn
    0.65
     boredom
    0.64
     Paraly
    0.64
     Haku
    0.63
     fate
    0.63
     sins
    0.62
    Activations Density 0.177%

    No Known Activations