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. 8-RES-JB
    6. 4467
    Prev
    Next
    INDEX
    Explanations

    paragraph separators that come in pairs

    oai_token-act-pair · gpt-3.5-turbo

    numerical values or dates within the text

    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
    •Partition Features Layer 8 GPT2 Small
    " Token" vs "token"
    •Examples of different Logit Weight Distributions - GPT2 Small Layer 8
    Partition Feature Example
    •Happy Hour
    Partition feature example

    No Comments

    Negative Logits
     Clause
    -0.94
     Coalition
    -0.87
     Labour
    -0.86
     Gul
    -0.85
     Par
    -0.84
     Mobil
    -0.83
     Polk
    -0.82
     Conc
    -0.81
     Ch
    -0.81
     Bill
    -0.81
    POSITIVE LOGITS
    nor
    1.32
    facebook
    1.22
    sorry
    1.19
    subject
    1.18
    cause
    1.18
    united
    1.17
    reply
    1.16
    john
    1.16
    else
    1.16
    negative
    1.15
    Activations Density 0.371%

    No Known Activations