© 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. 4723
    Prev
    Next
    INDEX
    Explanations

    mentions of physical impact or action, often involving forceful interactions like knocking

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "knock" and its variations

    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
     Complex
    -0.77
     Situation
    -0.76
    mberg
    -0.72
    heid
    -0.71
     Archives
    -0.70
     Municip
    -0.70
    rano
    -0.70
     Volunte
    -0.68
     Wikimedia
    -0.68
     Celeb
    -0.68
    POSITIVE LOGITS
     knock
    1.17
     knocks
    0.97
    down
    0.97
     knocking
    0.94
     knocked
    0.94
    kn
    0.85
    veyard
    0.85
    DOWN
    0.77
     unconscious
    0.76
     knockout
    0.75
    Activations Density 0.014%

    No Known Activations