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. 3-RES-JB
    6. 744
    Prev
    Next
    INDEX
    Explanations

    phrases related to protests or organized actions

    oai_token-act-pair · gpt-3.5-turbo

    phrases related to leaving or exiting situations

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 3-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    3
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    catentry
    -0.91
    Ħ¢
    -0.85
    ²¾
    -0.80
    İĭ
    -0.76
    pec
    -0.74
    ortunate
    -0.72
    eming
    -0.70
    nces
    -0.70
    itionally
    -0.68
    Ö¼
    -0.68
    POSITIVE LOGITS
    Redditor
    0.84
     heels
    0.74
     rooft
    0.73
     fence
    0.70
     jog
    0.69
     plank
    0.68
    ards
    0.67
     stairs
    0.66
     scrimmage
    0.66
     doors
    0.66
    Activations Density 0.141%

    No Known Activations