© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Initial Gemma-2B-IT SAEs
    5. Gemma-2B-IT
    7. Residual Stream
    9. 12-RES-JB
    11. 14657
    Prev
    Next
    INDEX
    Explanations

    contractions where an apostrophe is followed by a 't' for negation

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot

    It detects the trailing "’t"/"t" part of negative contractions (e.g., can't, don't, won't), signaling negation.

    oai_token-act-pair · gpt-5-miniTriggered by @phamvanluong2003
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    jbloom/Gemma-2b-IT-Residual-Stream-SAEs/gemma_2b_it_blocks.12.hook_resid_post_16384
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    HuggingFaceFW/fineweb
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.12.hook_resid_post
    Hook Layer
    12
    Architecture
    standard
    Context Size
    1,024
    Dataset
    Skylion007/openwebtext
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    878
    +0.14
    0.5%
    381
    +0.11
    0.4%
    1978
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    878
    +0.14
    0.06
    1124
    +0.11
    0.05
    1415
    +0.10
    0.03
    Negative Logits
     quelquefois
    -0.97
     Souha
    -0.96
     compréhen
    -0.92
     quoique
    -0.90
     Doña
    -0.88
     Alcalde
    -0.88
     endom
    -0.87
     sûrement
    -0.86
     parlamento
    -0.86
     Sinal
    -0.85
    POSITIVE LOGITS
     afford
    0.72
     consultato
    0.68
     immagine
    0.62
     autorytatywna
    0.61
    StructEnd
    0.60
     Normdatei
    0.59
    Cyfeiriadau
    0.59
     оригіналу
    0.59
     kaynağından
    0.59
    ngrx
    0.58
    Activations Density 0.121%

    No Known Activations