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

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsBlogSlackPrivacy & 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. 7-RES-JB
    6. 2345
    Prev
    Next
    INDEX
    Explanations

    instances where something is required or must be done

    oai_token-act-pair · gpt-3.5-turbo

    phrases related to obligations or requirements

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 7-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    7
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    ////////////////////////////////
    -0.83
    alf
    -0.77
    quickShipAvailable
    -0.75
    ergus
    -0.72
    calling
    -0.72
    fortunately
    -0.71
    fiction
    -0.70
    furt
    -0.69
    clips
    -0.69
    reddits
    -0.69
    POSITIVE LOGITS
     comply
    1.26
     obey
    1.24
     undergo
    1.22
     abide
    1.20
     perform
    1.17
     submit
    1.14
     disclose
    1.14
     undertake
    1.13
     attend
    1.13
     furnish
    1.11
    Activations Density 0.166%

    No Known Activations