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

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Open Source Sparse Autoencoders for all Residual Stream Layers of GPT2-Small
    5. GPT2-Small
    7. Residual Stream
    9. 3-RES-JB
    11. 23000
    Prev
    Next
    INDEX
    Explanations

    verbs indicating an action performed by a specific person

    oai_token-act-pair · gpt-3.5-turbo

    verbs that indicate actions or significant events in storytelling or narratives

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 3-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    3
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    respective
    -0.60
    ependent
    -0.54
    results
    -0.50
    ividually
    -0.50
    Compar
    -0.49
    necessary
    -0.49
    rina
    -0.48
    etermined
    -0.48
    OTAL
    -0.47
    irin
    -0.47
    POSITIVE LOGITS
    SourceFile
    0.72
     himself
    0.62
    bes
    0.55
     Himself
    0.54
    ĵĺ
    0.53
    Redditor
    0.51
     herself
    0.51
    isons
    0.49
     Paste
    0.49
     Vert
    0.49
    Activations Density 1.910%

    No Known Activations