© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26
    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 37-GEMMASCOPE-2-TRANSCODER-262K
    4. 121084
    Prev
    Next
    INDEX
    Explanations

    The neuron is associated with memory management, specifically the deallocation of dynamically allocated memory. The `MAX_ACTIVATING_TOKENS` list is dominated by the word `delete`, which is the keyword used in C++ and similar languages for this purpose. The `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` list shows common variables or structures that are targets of `delete`, such as `temp` (often a temporary pointer), `calculator` (an object pointer), `ptr` (a generic pointer), and `[]` (indicating an array allocated with `new[]`). The `TOP_ACTIVATING_TEXTS` further confirm this by showing code snippets involving `delete temp;`, `delete calculator;`, `delete xorResult;`, `delete ptr;`, `delete[] arr;`, and general discussions about memory leaks and deallocation in destructors.Therefore, the neuron detects or predicts the action of deleting dynamically allocated memory.Explanation: **deleting dynamically allocated memory**deleting dynamically allocated memory

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-27b-it/transcoder_all/layer_37_width_262k_l0_small_affine
    Prompts (Dashboard)
    238,145 prompts, 512 tokens each
    Dataset (Dashboard)
    lmsys + oasst1
    No Configuration Found
    Embeds
    IFrame
    Link
    Negative Logits
     वैदिक
    0.46
    зю
    0.46
     array
    0.45
     Array
    0.44
     arrays
    0.44
     multif
    0.42
     cycl
    0.42
    ائ
    0.42
     scattered
    0.41
    数组
    0.40
    POSITIVE LOGITS
     Schu
    0.41
    EventQueue
    0.37
    Snow
    0.37
     Washington
    0.35
    Nu
    0.35
     Wass
    0.35
    ashington
    0.34
    忌
    0.34
    缨
    0.34
    mes
    0.33
    Activations Density 0.000%

    No Known Activations