© 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. 24038
    Prev
    Next
    INDEX
    Explanations

    phrases related to forceful criticism or attacks

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "slam" and its variations, suggesting a focus on criticism or attack

    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
     Occupations
    -0.77
     Aware
    -0.70
    lements
    -0.68
     Revenue
    -0.65
     Expend
    -0.65
    lement
    -0.64
    ramer
    -0.63
     Discipline
    -0.63
    thus
    -0.62
    chn
    -0.62
    POSITIVE LOGITS
     dunk
    1.00
     slammed
    0.93
     slams
    0.90
     slamming
    0.86
     slam
    0.80
     shut
    0.80
     elbows
    0.78
    onite
    0.77
     smack
    0.76
    geon
    0.76
    Activations Density 0.015%

    No Known Activations