© 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. Gemma-3-12B-IT
    3. 12-GEMMASCOPE-2-RES-16K
    4. 5300
    Prev
    Next
    INDEX
    Explanations

    The neuron seems to detect or predict tokens related to programming contexts, particularly resource groups or project structures, often followed by a specific punctuation or delimiter. It also shows an inclination towards technical terms that might indicate names or types of resources.Looking at `MAX_ACTIVATING_TOKENS`:- `group` (appears 5 times)- `project` (appears 1 time)- `Case` (appears 1 time)Looking at `TOKENS_AFTER_MAX_ACTIVATING_TOKEN`:- `in` after `group`- `"` after `group`- `Table` after `Case`Looking at `TOP_ACTIVATING_TEXTS`:- Mentions `resource_group_name`- Mentions `resource group`- Mentions `project` (`git init my_project`)- Mentions `Test Case Table`Looking at `TOP_POSITIVE_LOGITS`:- `Algon` (might be abbreviation, part of something technical)- `eers` (like engineers, technical)- `methyl` (chemical/programming context)The most dominant pattern is `group` followed by `in` or `"` or other delimiters, and the presence of `resource_group` in the texts.resource group names

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-12b-it/resid_post/layer_12_width_16k_l0_medium
    Prompts (Dashboard)
    238,145 prompts, 512 tokens each
    Dataset (Dashboard)
    lmsys + oasst1
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    zi
    0.84
     luogo
    0.83
     numerosi
    0.82
     diminu
    0.82
     Năm
    0.80
    ziu
    0.78
     printemps
    0.77
    urunan
    0.76
     z
    0.75
     cambiamenti
    0.75
    POSITIVE LOGITS
    eous
    0.89
     Algon
    0.88
    smanship
    0.86
    জিৎ
    0.75
    ляется
    0.75
    eers
    0.74
     पोली
    0.73
    лдуу
    0.73
    ylmethyl
    0.73
    ्योपै
    0.73
    Activations Density 0.003%

    No Known Activations