© 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 · Open Source Sparse Autoencoders for all Residual Stream Layers of GPT2-Small
    5. GPT2-Small
    7. Residual Stream
    9. 0-RES-JB
    11. 23215
    Prev
    Next
    INDEX
    Explanations

    verbs or phrases indicating negative impact or harm

    oai_token-act-pair · gpt-3.5-turbo

    phrases related to causing harm or damage to credibility or stability

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 0-res-jb
    Configuration
    jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.hook_resid_pre
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    Data Type
    torch.float32
    Hook Point
    blocks.0.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    0
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     Fun
    -0.76
    active
    -0.70
     registered
    -0.69
     lucky
    -0.67
     Registered
    -0.67
    cal
    -0.67
    Fun
    -0.66
    iard
    -0.65
     Columb
    -0.64
    lication
    -0.62
    POSITIVE LOGITS
     undermine
    2.92
     undermined
    2.35
     undermines
    2.33
     undermining
    2.27
     undercut
    2.27
     weaken
    1.83
     undermin
    1.65
     discredit
    1.63
     sabotage
    1.57
     sabot
    1.56
    Activations Density 0.036%

    No Known Activations