© 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-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 66520
    Prev
    Next
    INDEX
    Explanations

    <ctrl94>thoughtThe user wants me to explain the behavior of a neuron based on provided lists.The explanation needs to be a concise phrase (3-20 words).I need to find a pattern in `MAX_ACTIVATING_TOKENS`, `TOKENS_AFTER_MAX_ACTIVATING_TOKEN`, `TOP_POSITIVE_LOGITS`, and `TOP_ACTIVATING_TEXTS`.Let's examine the lists:`MAX_ACTIVATING_TOKENS`:- 'out'- 'undo'- 'Touch'- 'page'- 'selected'- 'str'- 'gr'- 'VO'- 'Owner'- 'html'- 'est'- 'url'`TOKENS_AFTER_MAX_ACTIVATING_TOKEN`:- 'Menu' (after 'out')- 'Notification' (after 'undo')- 'Input' (after 'Touch')- 'Wrap' (after 'page')- 'Script' (after 'selected')- 'Collection' (after 'str')- 'Photo' (after 'gr', seems less direct)- 'Controller' (after 'VO', seems direct for VOController)- 'Group' (after 'Owner', seems direct for OwnerGroup)- 'html' (after 'html', not helpful)- 'Router' (after 'url')`TOP_POSITIVE_LOGITS`:- '='- '_'- 'will'- 'these'- 'u'- '?' (Thai character, likely ignorable or noise)- 'may'- '?' (Thai character)- 'h'- '?' (Chinese characters, likely ignorable or noise)`TOP_ACTIVATING_TEXTS`:- `outMenuWidth = $('.slideout-menu').width(); // toggle open class slideoutMenu.toggleClass("open"); // slide menu if (slideoutMenu.hasClass("` (starts with 'out')- `ndoObjectController -(void)awake FromNib { if (!_undoNotificationCenter) { _undoNotificationCenter = [[NSNotificationCenter alloc] init]; _undoNotificationQueue = [[NSNotificationQueue alloc] initWithNotification` (starts with 'undo')- `private static VirtualInput s_HardwareInput; static CrossPlatformInputManager() { s_TouchInput = new MobileInput(); s_HardwareInput = new StandaloneInput` (starts with 'Touch')- `= $("#page-wrap"), baseHeight = 0, $el; $pageWrap.height($pageWrap.height()); baseHeight = $pageWrap.height() - $mainContent` (contains 'page')- `= null; $scope.selectScript = function(script, index) { $scope.selectedScript = script; $scope.selectedRow = index; myAppFactory.updateTextArea(script` (contains 'selected')- `List<String> strCollection = new LinkedList<>(); for (MyClass item : myCollection) { strCollection` (starts with 'str')- `// cv::Mat bgrPhoto contains a test image cv::Mat3f hsvPhoto; bgrPhoto.convertTo(hsvPhoto, CV_32FC3); cv::cvtColor(hsvPhoto, hsv` (contains 'Photo', 'bgrPhoto', 'hsvPhoto'. Max activating token is 'gr'??<ctrl95>variable declarations

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     adolescentes
    -1.45
    xnn
    -1.37
     nettoyer
    -1.37
     pê
    -1.36
     tardes
    -1.35
    鐧
    -1.32
    strategie
    -1.29
     internos
    -1.27
     domingos
    -1.27
    dne
    -1.27
    POSITIVE LOGITS
     =
    1.71
    _
    1.62
     will
    1.48
     these
    1.44
    u
    1.39
    所以在
    1.38
     may
    1.36
    这也
    1.33
    h
    1.33
    了一些
    1.27
    Activations Density 0.009%

    No Known Activations