© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 16-GEMMASCOPE-2-RES-262K
    4. 13646
    Prev
    Next
    INDEX
    Explanations

    it's**Reasoning:**The `MAX_ACTIVATING_TOKENS` and `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` lists show common tokens like 'it', ',', 'a', 'is', 'or', and then words like 'the', 'carbon', 'jail', 'ever', 'too', 'primary', 'made', 'more'.The `TOP_ACTIVATING_TEXTS` provide context. Several examples seem to relate to the beginning of a sentence or clause, often starting with "it's" or discussing what "it" is or does.- "Anthrax **is** a deadly pathogen. **It's** a serious threat..."- "...This gas **is** valuable because:"- "...**it's** *created* through interaction..."- "...What **it** does:** The regex will match any string that: 1. Can start with any characters..."The phrase "it's" appears frequently as a starting point for discussions or descriptions. The `TOP_POSITIVE_LOGITS` do not seem to offer a coherent English pattern, but rather specific characters or jargon, suggesting the neuron might be sensitive to specific structural elements or beginnings of concepts, rather than semantic content itself. "it's" is a very common structural start to a descriptive clause or sentence

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-27b-it/resid_post/layer_16_width_262k_l0_medium
    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
     transgender
    0.55
     scuole
    0.54
     girls
    0.54
     clogging
    0.54
     teachers
    0.51
     clogged
    0.51
     corred
    0.50
     achie
    0.49
     פי
    0.49
     manis
    0.49
    POSITIVE LOGITS
    '
    0.47
    mailbox
    0.44
    iax
    0.43
    æg
    0.43
    စျေး
    0.43
    ذ
    0.42
    ตำแหน่ง
    0.42
    bmod
    0.42
    缈
    0.42
    兆
    0.41
    Activations Density 0.000%

    No Known Activations