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

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-12B-IT
    3. 12-GEMMASCOPE-2-RES-16K
    4. 5373
    Prev
    Next
    INDEX
    Explanations

    The neuron detects "the" followed by specific nouns or proper nouns. Looking at the `TOKENS_AFTER_MAX_ACTIVATING_TOKEN`, the common pattern is "the" followed by a capitalized word or a significant starting character.Let's analyze the provided lists to find the pattern:1. **`MAX_ACTIVATING_TOKENS`**: * `the` * `the` * `the` * `the` * `the` * `the` * `the` * `the` This strongly suggests the neuron is triggered by the presence of the word "the".2. **`TOKENS_AFTER_MAX_ACTIVATING_TOKEN`**: * `Lunar` * `Primary` * `O` * `sliding` * `"` * `refractive` * `**` * `GNU` This list shows what follows "the". We see capitalized words (`Lunar`, `Primary`, `GNU`, `O` which comes from Oort Cloud), a word starting with a quote (`"`) and a word starting with `**` (bold formatting in markdown). `sliding` and `refractive` are not capitalized but are followed by common noun phrases ("sliding window", "refractive index").The neuron detects "the" followed by a capitalized word

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-12b-it/resid_post/layer_12_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
    ラの
    1.13
    and
    1.04
    νο
    1.04
    вається
    1.02
    ようやく
    0.98
    ék
    0.94
    øm
    0.89
    ành
    0.89
    ڑک
    0.88
    án
    0.88
    POSITIVE LOGITS
     foothills
    1.09
     l
    1.05
     S
    0.97
     thunderstorms
    0.95
     использу
    0.95
    S
    0.94
     d
    0.94
     Nintendo
    0.93
     hurricanes
    0.93
    ます
    0.91
    Activations Density 0.095%

    No Known Activations