INDEX

Explanations

punctuation and conditional phrases

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

legal boilerplate

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

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

ins

-0.43

the

-0.42

-0.38

 painfully

-0.37

an

-0.36

Her

-0.35

 somehow

-0.35

 Schild

-0.35

PARTMENT

-0.35

POSITIVE LOGITS

 EconPapers

0.98

 Италијани

0.92

 незавершена

0.92

GEBURTSDATUM

0.90

 kasarigan

0.89

 Italijanski

0.89

хьтан

0.85

sizeCache

0.81

 Савезне

0.81

 ویکی‌پدی

0.79

Activations Density 0.018%

punctuation and conditional phrases

legal boilerplate

No Comments

No Known Activations