© 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. 17727
    Prev
    Next
    INDEX
    Explanations

    words related to physical destruction or incapacitation

    oai_token-act-pair · gpt-3.5-turbo

    occurrences of the word "crush" and its variants, typically related to physical damage or overwhelming situations

    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
    rog
    -0.74
    vernment
    -0.74
     appropri
    -0.74
    changes
    -0.72
    adapt
    -0.70
    ouf
    -0.70
    ativity
    -0.69
    nesota
    -0.67
    iator
    -0.67
    TERN
    -0.67
    POSITIVE LOGITS
     crush
    0.99
     Crush
    0.93
     crushing
    0.85
    avorite
    0.73
     earthqu
    0.73
     spoil
    0.72
    enance
    0.72
    meat
    0.71
    Boss
    0.71
     fracture
    0.71
    Activations Density 0.018%

    No Known Activations