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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Joseph Bloom · Open Source Sparse Autoencoders for all Residual Stream Layers of GPT2-Small
    3. GPT2-Small
    4. Residual Stream
    5. 3-RES-JB
    6. 13792
    Prev
    Next
    INDEX
    Explanations

    mentions of criminal convictions

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "convicted" in relation to criminal activity

    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
    arity
    -0.76
    patch
    -0.63
     anx
    -0.63
     RIS
    -0.62
    opic
    -0.62
    oos
    -0.61
    Bee
    -0.61
     Utt
    -0.60
    earable
    -0.60
    place
    -0.59
    POSITIVE LOGITS
     felon
    0.89
    icts
    0.86
     convicted
    0.85
     convict
    0.84
     guilty
    0.77
     sentenced
    0.76
     unfocusedRange
    0.76
    icted
    0.76
     debtor
    0.72
    iary
    0.72
    Activations Density 0.014%

    No Known Activations