© 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. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    5. Gemma-2-2B
    7. Residual Stream - 16k
    9. 21-GEMMASCOPE-RES-16K
    11. 8250
    Prev
    Next
    INDEX
    Explanations

    references to violence or aggression within contexts involving conflict or security threats

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-2B @ 21-gemmascope-res-16k
    Configuration
    google/gemma-scope-2b-pt-res/layer_21/width_16k/average_l0_70
    Prompts (Dashboard)
    36,864 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.21.hook_resid_post
    Hook Layer
    21
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     consenti
    -0.54
    allow
    -0.51
     accommodating
    -0.51
     allows
    -0.49
     allow
    -0.48
    Respond
    -0.47
     Allowing
    -0.47
     consenting
    -0.46
     allowing
    -0.46
    救命
    -0.45
    POSITIVE LOGITS
     invade
    0.98
     attack
    0.98
     stole
    0.97
     wre
    0.96
     hijack
    0.96
     attacking
    0.95
    attack
    0.93
     attacks
    0.91
     menyerang
    0.91
     steals
    0.91
    Activations Density 0.739%

    No Known Activations