© 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. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 24576
    9. 8-RES_FS24576-JB
    11. 10391
    Prev
    Next
    INDEX
    Explanations

    biographical information about historical figures

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs24576-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    Data Type
    torch.float32
    Hook Point
    blocks.8.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    8
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    314
    +0.13
    0.6%
    649
    +0.11
    0.5%
    465
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    465
    +0.13
    0.04
    475
    +0.11
    0.04
    663
    +0.10
    0.03
    Negative Logits
    agra
    -0.72
    sels
    -0.63
     leve
    -0.59
    gone
    -0.57
    akeru
    -0.56
     unemployment
    -0.55
     modelling
    -0.55
     thrott
    -0.54
     sequest
    -0.54
     mosqu
    -0.53
    POSITIVE LOGITS
    ]).
    1.15
    ],[
    1.12
    ]
    1.10
    ]"
    1.10
    ])
    1.06
    ]),
    1.05
    ][
    1.03
    ].
    0.98
    ]);
    0.96
    ]:
    0.96
    Activations Density 0.034%

    No Known Activations