© 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. Under Peer Review · Sparse Autoencoders for Pythia-70M-Deduped
    5. Pythia-70M-Deduped
    7. Residuals
    9. 4-RES-SM
    11. 13191
    Prev
    Next
    INDEX
    Explanations

    references to studies, methodologies, or data sets in academic or scientific writing

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    ctigges/pythia-70m-deduped__res-sm_processed/4-res-sm
    Prompts (Dashboard)
    32,768 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    32,768
    Data Type
    torch.float32
    Hook Name
    blocks.4.hook_resid_post
    Hook Layer
    4
    Architecture
    standard
    Context Size
    128
    Dataset
    EleutherAI/the_pile_deduplicated
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    21
    +0.26
    1.5%
    78
    +0.14
    0.8%
    136
    +0.11
    0.7%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    305
    +0.26
    0.03
    136
    +0.14
    0.00
    335
    +0.11
    0.02
    Negative Logits
    ities
    -1.46
     \"
    -1.34
     foreseeable
    -1.25
    rics
    -1.23
     execution
    -1.22
     licenses
    -1.20
    ership
    -1.17
     Site
    -1.16
     client
    -1.16
     accidental
    -1.16
    POSITIVE LOGITS
    ]).
    3.55
    ])
    3.15
    ]),
    3.15
    ].
    3.05
    ];
    2.94
    ],
    2.93
    ]);
    2.93
    )].
    2.91
    ].)
    2.83
    ]):
    2.79
    Activations Density 0.074%

    No Known Activations