Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Andy Arditi · Finding Misaligned Persona Features in Open-Weight Models
    3. Qwen2.5-7B-IT
    4. Resid Post - 131k
    5. 19-RESID-POST-AA
    6. 24708
    Prev
    Next
    INDEX
    Explanations

    requirement

    np_max-act-logits · gemini-2.0-flash

    text that signals the assistant persona—especially boilerplate AI self-descriptions, capability disclaimers, and generic greeting/help responses.

    oai_token-act-pair · gpt-5Triggered by @ariaw65

    tokens that are part of an AI assistant's generated response content.

    oai_token-act-pair · claude-4-5-haikuTriggered by @xiaoqingsun004
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    andyrdt/saes-qwen2.5-7b-instruct/resid_post_layer_19/trainer_1
    Dataset (Dashboard)
    Various
    Features
    131,072
    Data Type
    float32
    Hook Name
    blocks.19.hook_resid_post
    Architecture
    standard
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    disconnect
    -0.08
     İlk
    -0.07
     LN
    -0.07
     Dietary
    -0.07
     IntelliJ
    -0.07
     MainMenu
    -0.07
    ="#
    -0.07
     motors
    -0.07
    全线
    -0.07
    න
    -0.07
    POSITIVE LOGITS
     Schmidt
    0.08
    Locator
    0.07
    耋
    0.07
    OfFile
    0.07
     Polynomial
    0.07
    uat
    0.07
    芄
    0.07
    igsaw
    0.07
     emploi
    0.07
    _splits
    0.06
    Activations Density 36.103%

    No Known Activations