Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APISteerSAE EvalsBlog/PodcastNEWSlackPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlog/PodcastGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Open Source Sparse Autoencoders for all Residual Stream Layers of GPT2-Small
    3. GPT2-Small
    4. Residual Stream
    5. 5-RES-JB
    6. 4710
    Prev
    Next
    INDEX
    Explanations

    words related to physical force or destruction

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "crush" and its variations

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 5-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    5
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    iator
    -0.70
    rog
    -0.69
    iances
    -0.69
    landish
    -0.69
    ued
    -0.66
    adapt
    -0.65
    ouf
    -0.65
    ativity
    -0.65
    nesota
    -0.64
    vernment
    -0.64
    POSITIVE LOGITS
     crush
    0.84
     earthqu
    0.75
     Crush
    0.75
    Boss
    0.75
    dragon
    0.75
    meat
    0.73
     crushing
    0.70
    chu
    0.70
    DOWN
    0.68
    ulously
    0.68
    Activations Density 0.019%

    No Known Activations