© 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
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 93003
    Prev
    Next
    INDEX
    Explanations

    before Star Wars

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

    The neuron detects clauses built around the auxiliary “was” (or similar past‐tense auxiliaries) followed by a verb, flagging past‐progressive or passive constructions (“was expecting,” “was sworn in,” etc.).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     đang
    -1.23
     αὐτ
    -1.09
     are
    -0.99
     opět
    -0.98
     вони
    -0.96
     useRouter
    -0.95
    кли
    -0.95
     چربی
    -0.94
    Performances
    -0.93
    了一眼
    -0.92
    POSITIVE LOGITS
     last
    1.64
     when
    1.38
     years
    1.23
     each
    1.21
     was
    1.20
     initial
    1.20
     earlier
    1.19
     subsequent
    1.10
     previous
    1.09
     later
    1.07
    Activations Density 0.039%

    No Known Activations