© 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
    2. Qwen3-32B
    3. 32-RESID-BATCHTOPK-65K
    4. 51913
    Prev
    Next
    INDEX
    Explanations

    the impact, the most, the risks

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

    This neuron detects high-frequency function words, especially the definite article "the" and forms of the verb "to be" (e.g., "is", "are").

    oai_token-act-pair · gpt-5-miniTriggered by @yooniel31

    words that appear frequently at the beginning of sentences or after punctuation marks.

    oai_token-act-pair · claude-4-5-haikuTriggered by @yooniel31
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    adamkarvonen/qwen3-32b-saes/saes_Qwen_Qwen3-32B_batch_top_k/resid_post_layer_32
    Prompts (Dashboard)
    16,384 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     JP
    -0.05
    强åĬĽ
    -0.04
    è©
    -0.04
    åĩ¶æīĭ
    -0.04
    é¤IJæ¡Į
    -0.04
    åģ·åģ·
    -0.04
    å¼Ģçİ©ç¬ij
    -0.04
    abled
    -0.04
    æĬ¤èĤ¤
    -0.04
    蹦
    -0.04
    POSITIVE LOGITS
     $\
    0.05
     Arg
    0.04
     temper
    0.04
     TR
    0.04
    éĽĩ主
    0.04
     Wo
    0.04
     silently
    0.03
    IES
    0.03
    åĵªæĢķæĺ¯
    0.03
    /exp
    0.03
    Activations Density 41.204%

    No Known Activations