© 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. 34-GEMMASCOPE-RES-131K
    4. 25598
    Prev
    Next
    INDEX
    Explanations

    import com.eden.orchid.api.theme.assets.AssetPage import com.eden.orchid.api.theme.pages.OrchidReference import com.eden.orchid.api.resources.resource.StringResource` - Contains `Asset`, `Page` * `androidx.fragment.app.FragmentActivity; import com.kunzisoft.androidclearchroma.ChromaDialog; import com.kunzisoft.androidclearchroma.IndicatorMode; import com.kunzisoft` - Contains `Dialog` (implied by ChromaDialog) * `databind.ObjectMapper; import com.twilio.Twilio; import com.twilio.converter.DateConverter; import com.twilio.converter.Promoter; import com.twilio.exception.Tw` - Contains `Date`, `Converter` * `cpw.mods.fml.relauncher.SideOnly; import mods.eln.misc.Direction; import net.minecraft.client.gui.GuiScreen; import net.minecraft.entity.player.` - Contains `Direction` * `net.minecraftforge.server.permission.DefaultPermissionLevel; import net.minecraftforge.server.permission.PermissionAPI` - Contains `Permission` * `diff --git a/chrome/android/java/src/org/chromium/chrome/browser/notifications/NotificationWrapperBuilderFactory.java b/chrome/android/java/src/org/chromium/chrome/browser/notifications/` - Contains `Notification`Observations:- `MAX_ACTIVATING_TOKENS` are common programming terms, often related to modules, data types, or functionality within software (App, DB, Chat, Mail, Asset, Date, Direction, Permission, Notification).- `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` are frequently utility-like suffixes or common programming terms (`View`, `Table`, `Util`, `Page`, `Converter`, `API`, `Wrapper`).- `TOP_POSITIVE_LOGITS` are predominantly semicolons and other punctuation often found at the end of code lines or structures.- The `TOP_ACTIVATING_TEXTS` show these tokens appearing within `import` statements or code prefixes/suffixes.The neuron seems to be triggered by common programming identifiers, particularly those that are followed by common programming suffixes or followed by punctuation like semicolons, often found within import statements or code definitions.The pattern is identifying common programming entities/modules when followed by typical code constructs or terminators.Let's try to summarize this:- "programming entities followed by separators"- "code entities with terminators"- "software modules and their attributes"- "common program components and code structure"Considering the "TOP_POSITIVE_LOGITS" being mostly ";", this suggests a strong association with code termination or separation. The "MAX_ACTIVATING_TOKENS" are often nouns representing parts of a system.A good concise phrase could be: "common programming identifiers followed by code separators" or "program components and code structure".Let's simplify. The most striking pattern is the program entity followed by `;` or similar code structuring elements."code components followed by terminators" fits well. code components followed by terminators

    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_34/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
     poichè
    -1.22
    いえ
    -1.21
     manifiesta
    -1.20
    pherd
    -1.16
     there
    -1.14
    DrawerToggle
    -1.14
     bezw
    -1.13
    
    -1.13
     esconde
    -1.13
    Moreover
    -1.09
    POSITIVE LOGITS
    ;
    3.55
    ;
    
    2.45
    ;</
    2.03
    }$;
    1.98
    $;
    1.83
    ];
    1.79
    ;");
    1.75
    %;
    1.74
    ;
    1.70
    ;}
    1.66
    Activations Density 0.032%

    No Known Activations