© 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. 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. 4415
    Prev
    Next
    INDEX
    Explanations

    references to public figures, particularly George Osborne

    oai_token-act-pair · gpt-3.5-turbo

    mentions of political figures, particularly George Osborne

    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
    LESS
    -0.79
    alde
    -0.73
    uay
    -0.72
    hips
    -0.72
    UE
    -0.66
    ###
    -0.65
    +++
    -0.65
    ging
    -0.64
    ivity
    -0.64
    essors
    -0.64
    POSITIVE LOGITS
     Osborne
    1.01
    £
    0.86
     Lans
    0.77
    forth
    0.74
    artney
    0.70
    ©¶æ¥µ
    0.70
    boro
    0.69
     austerity
    0.68
    mega
    0.66
    mania
    0.66
    Activations Density 0.008%

    No Known Activations