© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Joseph Bloom · Initial Gemma-2B SAEs
    3. Gemma-2B
    4. Residual Stream
    5. 10-RES-JB
    6. 1611
    Prev
    Next
    INDEX
    Explanations

    sentences that begin with "how to" or similar instruction phrases

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    jbloom/Gemma-2b-Residual-Stream-SAEs/gemma_2b_blocks.10.hook_resid_post_16384
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    chanind/openwebtext-gemma
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.10.hook_resid_post
    Hook Layer
    10
    Architecture
    standard
    Context Size
    1,024
    Dataset
    ctigges/openwebtext-gemma-1024-cl
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    50
    +0.44
    2.4%
    2004
    +0.10
    0.5%
    1141
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    1372
    +0.44
    0.03
    1141
    +0.10
    0.03
    2004
    +0.10
    0.03
    Negative Logits
    <bos>
    -2.27
    /***
    
    -0.55
    <?
    -0.53
    enting
    -0.53
    estab
    -0.49
    ždý
    -0.49
    const
    -0.49
    ⓧ
    -0.49
    int
    -0.48
    linker
    -0.48
    POSITIVE LOGITS
     Minang
    1.20
     sovere
    1.16
     philanth
    1.16
     Jambi
    1.16
     practition
    1.13
     unlaw
    1.11
     bandung
    1.10
     Palembang
    1.09
     Juf
    1.08
     Karang
    1.07
    Activations Density 0.070%

    No Known Activations