© 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. 38-GEMMASCOPE-2-TRANSCODER-262K
    4. 753
    Prev
    Next
    INDEX
    Explanations

    **TOP_POSITIVE_LOGITS**:`give`, `Give`, `Give`, `give`, `給`, `giving`, `GIVE`, `给`, `geven`, `Giving`This list overwhelmingly points to the concept of "giving" or "to give".**MAX_ACTIVATING_TOKENS**:`you`, `you`, `**`, `a`, `coworker`, `Give`, `get`, `preferential`, `anyone`This list includes "Give" and "preferential". "Coworker" might also relate to an interaction where giving is involved.**TOKENS_AFTER_MAX_ACTIVATING_TOKEN**:`think`, `think`, `side`, `*`, `a`, `the`, `some`, `treatment`, `else`When "preferential" is followed by "treatment", it strongly reinforces a specific meaning.**TOP_ACTIVATING_TEXTS**:The texts mention:- "give your coworker"- "giving preferential treatment"- "Give the raccoon space"- "give some credit"- "giving preferential treatment"**Pattern Identification**:The neuron seems strongly associated with the act of "giving". The 'TOP_POSITIVE_LOGITS' are almost exclusively variations of "give". The 'MAX_ACTIVATING_TOKENS' include instances of "Give" and "preferential", and the 'TOKENS_AFTER_MAX_ACTIVATING_TOKEN' sometimes show "treatment" after "preferential". The 'TOP_ACTIVATING_TEXTS' provide concrete examples like "give your coworker", "giving preferential treatment", all revolving around the concept of bestowal or granting.The most specific and recurring theme is "giving preferential treatment" or the general act of "giving".**Constraint Checklist & Confidence Score**:1. Concise explanation (3-20 words): Yes.2. Captures what neuron detects/predicts: Yes.3. Finds patterns in lists: Yes.4. Explanation is a word/phrase/pattern: Yes.5. About tokens following/preceding: Yes, implicitly through patterns.6. About words starting with a sequence: Not explicitly, but implied.7. Avoid listing all tokens: Yes.8. Find patterns, don't just list: Yes.9.**giving preferential treatment**

    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_38_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
     ecology
    0.40
    λι
    0.35
    噜
    0.34
    生态
    0.34
    蛍
    0.34
     carpet
    0.33
     npc
    0.33
    ecology
    0.33
     lẫn
    0.33
     полити
    0.33
    POSITIVE LOGITS
     give
    1.69
    Give
    1.66
     Give
    1.63
    give
    1.59
    給
    1.46
     giving
    1.45
     GIVE
    1.43
    给
    1.42
     geven
    1.39
     Giving
    1.36
    Activations Density 0.143%

    No Known Activations