© 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. 3-GEMMASCOPE-2-TRANSCODER-262K
    4. 44512
    Prev
    Next
    INDEX
    Explanations

    **TOP_ACTIVATING_TEXTS**:* `StaminaBar = script.Parent.ScreenGui.StaminaFrame.Size -- StaminaBar = UDim2.new(Stamina / StaminaCapacity, 0, 1, 0) -- script` * This contains `UDim2.new`, which relates to GUI sizing. `Dim2` is a type of dimension.* `("GuiService"):GetScreenGui() local frame = Instance.new("Frame") frame.Size = UDim2.new(1, 0, 1, 0) frame.BackgroundTransparency = 1` * Another instance of `UDim2.new`.* `**Difficulty/Cost:** DIY / Low-Medium (Free, potentially time-consuming) * **BIOS/UEFI Settings:** Some BIOS settings can affect power management. * **Troubleshooting:** 1. **Reset` * Contains `Settings`.* `This is the most common and well-supported method. **1. Enable IOMMU in BIOS/UEFI:** * Reboot your computer and enter the BIOS/UEFI setup. * Look for settings related` * Contains `Settings`.* `Starts to initialize, but doesn't yet have an IP address. * **Boot ROM/BIOS/UEFI:** Executes initial instructions stored in read-only memory. * Key Actions: * **POST` * Contains `Settings` implicitly in `BIOS/UEFI`.* `(Hard Drive, SSD): Long-term data storage. 2. **Firmware/BIOS/UEFI:** (Very close to the hardware) * Low-level software embedded in hardware. Initializes hardware and` * Contains `Settings` implicitly in `BIOS/UEFI`.* `* Boot your computer from the Gentoo installation medium (USB/CD). You might need to adjust your BIOS/UEFI settings to change the boot order. * You should see a boot prompt. Type boot and press Enter` * Contains `settings`.* `) is *required* for GPU passthrough. Check your motherboard's documentation. * **BIOS/UEFI:** Enable IOMMU/VT-d/AMD-Vi in your BIOS/UEFI settings. Also` * Contains `settings`.* `letter. Right-click on the partition and select "Change Drive Letter and Paths..." * **BIOS/UEFI Check:** Restart your computer and enter the` * Contains `Settings` implicitly in `BIOS/UEFI`.**Pattern Identification**:* The `MAX_ACTIVATING_TOKENS` list shows `Dim` and `FI`.* The `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` shows `2` after `Dim`, and `Settings`, `settings`, `:**`, `Check` after `FI`.* The `TOP_ACTIVATING_TEXTS` frequently mention `Settings` or `settings` in contexts like `BIOS/UEFI Settings`.* Specifically, the presence of `UDim2.new` in the text lists, which includes `Dim2`, suggests a connection to dimensions or sizing, often found in GUI contexts. The `FI` tokens followed by `Settings` are also a strong signal.Combining these, the neuron seems to be triggered by specific configurations or settings, especially those related to dimensions (like `UDim2`) or system settings (like `BIOS/UEFI Settings`). The 'FI' token appears to be a strong precursor to 'Settings'.Let's re-evaluate the prompt requirements:- Conciseness (3-20 words)- Single phrase- "finds patterns in lists"- "about tokens following or preceding certain tokens"- "about words starting with a sequence"- "do not mention 'tokens' or 'patterns'"- "do not capitalize the first letter unless it is a proper noun"- "specific"- "majority should"The neuron activates for 'Dim' and 'FI'.'FI' is strongly followed by 'Settings' (or 'settings').The texts mention 'BIOS/UEFI Settings', 'UDim2'.A possible explanation could be related to 'settings' or 'configuration'.The 'Dim' token with '2' after it is a bit disconnected from the 'Settings' theme.However, 'FI' + 'Settings' is strong. The `UDim2` also hints at configuration/settings.Let's focus on `FI` followed by `Settings`.The `TOP_POSITIVE_LOGITS` are diverse and don't help pinpoint a specific word pattern easily.The neuron activates on `Dim` and `FI`.After `FI`, it sees `Settings`, `:**`, `settings`, `Check`.The texts contain `BIOS/UEFI Settings`, `UDim2.new`.This suggests a neuron that might be looking for configuration parameters, system settings, or visual layout dimensions.Let's try to make it specific. If the neuron is sensitive to sequences that look like they are part of a configuration or settings block:- `FI` -> `Settings`- `UDim2` (from text) -> related to dimensions/settingsConsider the `Dim` activations. If it's `Dim` followed by `2`, maybe it's about numeric settings or dimensions?The `TOP_POSITIVE_LOGITS` are very varied and cryptic.Let's revisit the core prompt: "what the neuron detects or predicts by finding patterns in lists."The pattern is `FI` preceding `Settings`/`settings`. The `Dim` is also a trigger.The texts provide context: `UDim2.new` (dimensions/GUI), `BIOS/UEFI Settings`.Possible explanations:- `FI..Settings`- `UDim2` related- configuration parameters- system settings"FI settings" is too short."BIOS/UEFI settings" is too specific to the text, not the token lists."Configuration settings" covers `UDim2` and `Settings`.Let's look at `Dim` + `2`. Could it be related to `UDim2.new(value, 0, value, 0)` where the first `value` is important?The text `StaminaBar = script.Parent.ScreenGui.StaminaFrame.Size -- StaminaBar = UDim2.new(Stamina / StaminaCapacity, 0, 1, 0)` shows `UDim2.new(Stamina / StaminaCapacity, 0, 1, 0)`.Here the first value `Stamina / StaminaCapacity` is a variable, and `1` is a fixed value.The neuron fires on `Dim` and `FI`.After `FI`, it sees `Settings` or `settings`.This strongly points to a neuron that recognizes sequences related to settings or configurations.A concise explanation: "FI followed by settings" - this is 4 words.Or "configuration settings" - 2 words. Too short.Let's check the `MAX_ACTIVATING_TOKENS` again: `Dim`, `Dim`, `Dim`, `FI`, `FI`, `FI`, `FI`, `FI`, `FI`, `FI`.And `TOKENS_AFTER_MAX_ACTIVATING_TOKEN`: `2`, `2`, `2`, `Settings`, `:**`,

    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_3_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
    itt
    0.75
    man
    0.74
    ing
    0.71
    og
    0.71
    nia
    0.70
    o
    0.70
    erv
    0.69
    ension
    0.69
    I
    0.69
    ub
    0.68
    POSITIVE LOGITS
    ле
    0.99
    0
    0.94
    ের
    0.89
     erano
    0.84
     взрос
    0.82
    ות
    0.81
     auront
    0.79
    ڈ
    0.79
    )'
    0.79
    の
    0.79
    Activations Density 0.003%

    No Known Activations