© 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 EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Apollo Research · Taylor · Identifying Functionally Important Features with End-to-End Sparse Dictionary Learning
    3. GPT2-Small
    4. Residual Similar CE E2E Recon
    5. 6-RES_SCEFR-AJT
    6. 2452
    Prev
    Next
    INDEX
    Explanations

    future tense verbs related to decisions, actions, or potential outcomes

    oai_token-act-pair · gpt-3.5-turbo

    modal auxiliary verbs as well as their variations, particularly focusing on the forms of "will" and "would."

    oai_token-act-pair · gpt-3.5-turboTriggered by @danbraun
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    neuronpedia/gpt2-small__res_scefr-ajt/6-res_scefr-ajt
    Prompts (Dashboard)
    12,288 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    46,080
    Data Type
    torch.float32
    Hook Point
    blocks.6.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    apollo-research/Skylion007-openwebtext-tokenizer-gpt2
    Hook Point Layer
    6
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    367
    +0.15
    0.6%
    339
    +0.14
    0.5%
    551
    +0.14
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    551
    +0.15
    0.15
    703
    +0.14
    0.11
    718
    +0.14
    0.11
    Negative Logits
    usage
    -0.86
    eyed
    -0.85
    noticed
    -0.80
    living
    -0.80
    treated
    -0.77
    lee
    -0.77
    gdala
    -0.76
     visiting
    -0.76
    essor
    -0.76
     inspecting
    -0.75
    POSITIVE LOGITS
     suffice
    1.47
     coincide
    1.34
     be
    1.31
     imply
    1.28
     entail
    1.27
     undoubtedly
    1.27
     doubtless
    1.22
     consist
    1.19
     probably
    1.19
     likely
    1.14
    Activations Density 1.423%

    No Known Activations