© 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 EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Andy Arditi · Finding Misaligned Persona Features in Open-Weight Models
    5. Qwen2.5-7B-IT
    7. Resid Post - 131k
    9. 11-RESID-POST-AA
    11. 68574
    Prev
    Next
    INDEX
    Explanations

    Times

    np_max-act · gemini-2.0-flash

    The neuron detects the "assistant" speaker/role token (i.e., the assistant turn marker) in chat-style transcripts.

    oai_token-act-pair · gpt-5-miniTriggered by @vetterc0
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    andyrdt/saes-qwen2.5-7b-instruct/resid_post_layer_11/trainer_1
    Dataset (Dashboard)
    Various
    Features
    131,072
    Data Type
    float32
    Hook Name
    blocks.11.hook_resid_post
    Architecture
    standard
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    .Dao
    -0.08
     tornado
    -0.07
    確認
    -0.07
    臻
    -0.07
    棰
    -0.07
    也比较
    -0.07
    铩
    -0.07
    ㊗
    -0.07
    比较大
    -0.07
     �
    -0.06
    POSITIVE LOGITS
     hates
    0.07
    ////////////////
    0.07
    pcs
    0.06
    Analytics
    0.06
     üzerinde
    0.06
    军队
    0.06
     Multi
    0.06
    arity
    0.06
     recip
    0.06
     인간
    0.06
    Activations Density 0.023%

    No Known Activations