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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Gemma-3-4B-IT
    3. 22-GEMMASCOPE-2-RES-16K
    4. 3759
    Prev
    Next
    INDEX
    Explanations

    evil AI, monsters

    np_acts-logits-general · gemini-2.5-flash-lite

    The neuron fires strongly on genre labels—especially darker ones—such as “Horror,” “Mystery,” “dark,” or “scary.”

    oai_token-act-pair · o4-miniTriggered by @drejcpesjakpesjak
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-4b-it/resid_post/layer_22_width_16k_l0_medium
    Prompts (Dashboard)
    238,145 prompts, 512 tokens each
    Dataset (Dashboard)
    lmsys + oasst1
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     sunshine
    0.91
     নির্বাচ
    0.88
     कोहली
    0.86
     Barça
    0.85
     sparkles
    0.84
     entrepreneur
    0.84
     ధ
    0.83
    毽
    0.82
    🏀
    0.80
     Spark
    0.80
    POSITIVE LOGITS
     creepy
    2.34
     horror
    2.30
     Horror
    2.26
     spooky
    2.23
     sinister
    2.20
     eerie
    2.15
     terrifying
    2.13
     horrors
    2.08
     demonic
    2.07
     horrifying
    2.04
    Activations Density 0.919%

    No Known Activations