INDEX

Explanations

humorous expressions and statements that convey seriousness or skepticism

oai_token-act-pair · gpt-4o-mini Triggered by @bot

transition to serious point

np_acts-logits-general · gemini-2.5-flash-lite

The neuron detects discourse markers and tone-shifters—words/phrases that signal emphasis, contrast, or a change in register (e.g., "seriously", "on a more helpful note", "despite", "however").

oai_token-act-pair · gpt-5-mini Triggered by @vetterc0

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 ویکی‌پدی

-0.90

IsMutable

-0.84

AutoField

-0.79

 estekak

-0.77

httphttps

-0.75

 المعيارى

-0.72

CppMethod

-0.71

 tartalomajánló

-0.71

setVerticalGroup

-0.69

SharedCtor

-0.69

POSITIVE LOGITS

Seriously

0.45

 seriously

0.40

 Seriously

0.38

 FACT

0.33

 serious

0.32

 really

0.32

 thực

0.32

 sensible

0.31

 fact

0.31

 серь

0.31

Activations Density 0.041%

humorous expressions and statements that convey seriousness or skepticism

transition to serious point

The neuron detects discourse markers and tone-shifters—words/phrases that signal emphasis, contrast, or a change in register (e.g., "seriously", "on a more helpful note", "despite", "however").

No Comments

No Known Activations