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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Andy Arditi · Finding Misaligned Persona Features in Open-Weight Models
    3. Llama3.1-8B-IT (Instruct)
    4. Resid Post - 131k
    5. 11-RESID-POST-AA
    6. 125690
    Prev
    Next
    INDEX
    Explanations

    if

    np_max-act · gemini-2.0-flash

    This neuron detects apologetic or self-deprecating hedges (e.g. “forgive me if …,” “sorry if …,” “excuse me if …”).

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    andyrdt/saes-llama-3.1-8b-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
     двух
    -0.07
     bulld
    -0.07
     Crest
    -0.06
    -link
    -0.06
     assaulting
    -0.06
     dirty
    -0.06
     guideline
    -0.06
     plummet
    -0.06
     working
    -0.06
     problem
    -0.06
    POSITIVE LOGITS
    屋
    0.07
    ."'";↵
    0.07
    уча
    0.06
    zione
    0.06
    읍
    0.06
    _HS
    0.06
    ีป
    0.06
    ючи
    0.06
     ณ
    0.06
     철
    0.06
    Activations Density 0.015%

    No Known Activations