© 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. Llama3.1-8B-IT
    7. Resid Post - 131k
    9. 11-RESID-POST-AA
    11. 57126
    Prev
    Next
    INDEX
    Explanations

    conflict/negative situations

    np_max-act · gemini-2.0-flash

    The neuron is primarily responding to common function words (e.g. “never,” “to,” “do,” “anything,” “about”)—in other words, it detects high-frequency English stopwords.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8

    finds procedural/explanatory or administrative text—sentences that state policies, instructions, processes, or formal reports.

    oai_token-act-pair · gpt-5-miniTriggered by @vetterc0
    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
     nhất
    -0.07
     �
    -0.06
     Angela
    -0.06
     Flatten
    -0.06
     Sheikh
    -0.06
     verifies
    -0.06
     cet
    -0.06
     Wa
    -0.06
     fark
    -0.06
     tempfile
    -0.06
    POSITIVE LOGITS
    upported
    0.07
    FUL
    0.07
    ید
    0.07
    completion
    0.06
     сю
    0.06
    ază
    0.06
    رض
    0.06
    parsers
    0.06
                
    0.06
     bunch
    0.06
    Activations Density 0.095%

    No Known Activations