© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 31-GEMMASCOPE-2-RES-262K
    4. 92867
    Prev
    Next
    INDEX
    Explanations

    specificThe examples show a mix of introductory phrases and technical/code snippets. They often start with specific tokens like "specific," "It," "Punk," "indeed," "transcribe," or code markers (`<?php`, ````python`, ````c++`) followed by some context. The positive logits suggest words like "begint", "retry", "pathogenicity", "tenacity", "told", "or", "lan", many of which seem somewhat unrelated or general.Considering the MAX_ACTIVATING_TOKENS and TOP_ACTIVATING_TEXTS:- "specific" is a token itself.- "Punk" is a token, and "Punk's Acceleration" appears in texts.- "indeed" appears in texts.- "transcribe" appears in texts related to whisper.- The code snippets are distinct.- The texts discuss specific topics like RolePlay, AI in games, Punk, bowling, tinnitus, Prime numbers, and code.The common thread seems to be introducing diverse, specific topics or concepts, often with a declarative or descriptive tone. The presence of code snippets and technical terms like "transcribe" and "AI" suggests a neuron that recognizes structured or specialized language.The phrase "specific definitions or introductions" seems to fit."specific definitions or introductions" has 4 words. It's concise and captures the essence of introducing specific topics, definitions, or concepts, as seen in the texts and the initial tokens. specific definitions or introductions

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-27b-it/resid_post/layer_31_width_262k_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
    ələ
    0.53
     pada
    0.48
     של
    0.47
    ದಲ್ಲಿ
    0.47
    یل
    0.47
     شی
    0.46
    ोच्च
    0.46
     όπου
    0.46
    ânt
    0.46
    asional
    0.46
    POSITIVE LOGITS
     begint
    0.47
    retry
    0.47
     pathogenicity
    0.45
    to
    0.45
    er
    0.44
     tenacity
    0.44
    told
    0.43
    or
    0.43
    គុ
    0.41
    lan
    0.41
    Activations Density 0.001%

    No Known Activations