© 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. 10-RES-JB
    11. 6915
    Prev
    Next
    INDEX
    Explanations

    instances where something is being explained, described, or detailed

    oai_token-act-pair · gpt-3.5-turbo

    instances of the phrase "lay out" and its variations, indicating the presentation of plans or arguments

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 10-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    10
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    tes
    -0.68
    blers
    -0.68
    enf
    -0.65
    Bir
    -0.65
    asion
    -0.63
    oso
    -0.62
    WT
    -0.61
    ONG
    -0.61
    Redditor
    -0.60
     luck
    -0.60
    POSITIVE LOGITS
     plans
    1.18
     principles
    1.14
     outlines
    1.10
     reasons
    1.09
     criteria
    1.07
     objectives
    1.06
     guidelines
    1.05
     strategies
    1.05
     goals
    1.04
     how
    1.04
    Activations Density 0.186%

    No Known Activations