© 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. 10-GEMMASCOPE-RES-131K
    7. 75364
    Prev
    Next
    INDEX
    Explanations

    ha sequence detection

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

    syllables or letter combinations "ha" in various names and words, particularly in non-English or technical terms.

    oai_token-act-pair · claude-3-7-sonnet-20250219Triggered by @neilrathi

    The neuron strongly activates on the subtoken "ha" (i.e. occurrences of the character sequence “ha”) in words.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_10/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
    .”
    -2.78
    
    -2.77
     obter
    -2.64
    焌
    -2.55
     peuvent
    -2.50
    ſelves
    -2.50
    With
    -2.47
    !”
    -2.42
     すぐ
    -2.30
    鐐
    -2.30
    POSITIVE LOGITS
    ha
    2.81
    "
    2.58
    了一个
    2.50
    
    2.45
    岽
    2.44
    ..?
    2.22
     ſame
    2.22
    bk
    2.22
     throng
    2.20
    ꦶ
    2.14
    Activations Density 0.019%

    No Known Activations