© 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 EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 37-GEMMASCOPE-2-TRANSCODER-262K
    4. 204564
    Prev
    Next
    INDEX
    Explanations

    The neuron detects patterns related to the diagnosis and characteristics of Borderline Personality Disorder (BPD) and Antisocial Personality Disorder (ASPD), particularly their behavioral and emotional dysregulation aspects.**Evidence from lists:*** **TOP_POSITIVE_LOGITS:** `patrick` (potentially related to character names in examples), `personality`, `pacific` (could be related to 'disorder' or part of a specific type), `acted` (behavior).* **TOP_ACTIVATING_TEXTS:** The texts heavily feature discussions about "Borderline Personality Disorder (BPD)" and "Antisocial Personality Disorder (ASPD)". They also describe core symptoms like "behavior", "fear of abandonment", "emotional dysregulation", "impulsivity", and "regression".* **MAX_ACTIVATING_TOKENS:** `B` (likely for BPD), `ASP` (for ASPD).* **TOKENS_AFTER_MAX_ACTIVATING_TOKEN:** `PD` (following `B` -> `BPD`), `fears`, `impuls` (impulsivity), `reg` (regression), `behavior`.**Synthesizing the pattern:**The neuron seems to activate when it sees abbreviations or mentions related to psychological disorders (BPD, ASPD) and their associated behavioral/emotional traits. The tokens following `B` (`PD`) strongly suggest BPD. The presence of `ASP` and tokens like `impuls`, `reg`, `fears`, `behavior` points to the diagnostic criteria of these personality disorders.Explanation should be concise (3-20 words).Candidate explanations:- BPD, ASPD, and associated behaviors- Personality disorder traits and behaviors- Borderline Personality Disorder and ASPD- Behavioral dysregulation in personality disorders

    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
    Not in Any Lists

    No Comments

    Negative Logits
     ablation
    0.40
     недостатки
    0.39
     collo
    0.37
     கோடி
    0.37
     playgrounds
    0.36
     flaw
    0.36
     aligned
    0.35
    塙
    0.34
     brach
    0.33
     JOptionPane
    0.33
    POSITIVE LOGITS
    patrick
    0.42
    ॉर्क
    0.41
    व्यक्तित्व
    0.39
    +'/
    0.39
     Поль
    0.38
    FFE
    0.38
    磵
    0.38
    pacific
    0.38
     पॉलिटी
    0.37
     acted
    0.37
    Activations Density 0.003%

    No Known Activations