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

    Neuronpedia

    APIAssistant AxisNEWCircuit 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. 10-RES-JB
    6. 19999
    Prev
    Next
    INDEX
    Explanations

    words related to fantasy, gaming, and elemental powers, most notably focusing on fire, darkness, and summoning abilities

    oai_token-act-pair · gpt-3.5-turbo

    references to elements from fantasy or RPG-themed content

    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
     supervised
    -0.89
     affordable
    -0.84
     includ
    -0.80
     duly
    -0.78
     recreational
    -0.76
     elig
    -0.75
     financially
    -0.75
     federally
    -0.74
     charitable
    -0.73
     antib
    -0.73
    POSITIVE LOGITS
     Sword
    1.32
     Breaker
    1.32
    blade
    1.28
    bringer
    1.24
     Swords
    1.24
     Dragon
    1.23
     Scythe
    1.22
    steel
    1.20
    wolf
    1.20
    sword
    1.19
    Activations Density 0.167%

    No Known Activations