© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports 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. 0-RES-JB
    11. 22290
    Prev
    Next
    INDEX
    Explanations

    negative actions or behaviors related to attacks on reputation, characterized by words like "smear," "slander," "defamation," and "distortions."

    oai_token-act-pair · gpt-3.5-turbo

    terms associated with attacks on reputation and character

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 0-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    0
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    jo
    -0.78
    iration
    -0.74
     Wond
    -0.74
     Ele
    -0.74
     autom
    -0.73
    hook
    -0.72
     Happiness
    -0.72
    HT
    -0.72
     aw
    -0.69
     Zen
    -0.68
    POSITIVE LOGITS
     smear
    3.21
     slander
    1.77
     libel
    1.73
     defamation
    1.72
     disinformation
    1.62
     misinformation
    1.61
     distort
    1.60
     wedge
    1.55
     distortion
    1.53
     distortions
    1.51
    Activations Density 0.051%

    No Known Activations