© 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. 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. 16589
    Prev
    Next
    INDEX
    Explanations

    instances where the phrase "go ahead" is used

    oai_token-act-pair · gpt-3.5-turbo

    phrases that indicate making progress or taking action

    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
    mini
    -0.84
    rug
    -0.76
    itar
    -0.72
    ondo
    -0.69
    cu
    -0.67
    odor
    -0.67
    lee
    -0.66
    VID
    -0.65
    ASED
    -0.65
    ifted
    -0.64
    POSITIVE LOGITS
     ahead
    0.80
    erk
    0.78
    olicy
    0.72
    manship
    0.71
     Dragonbound
    0.71
    eering
    0.67
    icter
    0.67
     fert
    0.66
     showc
    0.66
    ģĸ
    0.65
    Activations Density 0.008%

    No Known Activations