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. 8-RES-JB
    6. 3766
    Prev
    Next
    INDEX
    Explanations

    words related to physical structures or objects, specifically focusing on the condition "sh" with values primarily at 8 or 9

    oai_token-act-pair · gpt-3.5-turbo

    references to "sh" as it appears in various contexts, potentially indicating a focus on a particular sound or aspect related to that string

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    8
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     numbered
    -0.79
    STER
    -0.67
     Ajax
    -0.64
     Leilan
    -0.60
     NEO
    -0.60
     dating
    -0.60
     Luther
    -0.59
     quiz
    -0.59
     Nazis
    -0.57
     Claud
    -0.56
    POSITIVE LOGITS
    ifts
    1.17
    tml
    1.16
    ackle
    1.15
    ifting
    1.14
    uffle
    1.11
    IFT
    1.09
    aft
    1.07
    adows
    1.06
    nikov
    1.04
    aded
    1.04
    Activations Density 0.025%

    No Known Activations