INDEX

Explanations

tokens that consist of underscores or related patterns in data and formatting

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

Gravfile names and greek math notation

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

elakaan

-0.28

 terima

-0.25

istungs

-0.24

 sánchez

-0.24

shutil

-0.24

 yöne

-0.24

 наси

-0.24

utebol

-0.24

dAtA

-0.23

 DataTypes

-0.23

POSITIVE LOGITS

ValueStyle

0.92

Билгалдахарш

0.86

httphttps

0.78

 kasarigan

0.73

 nakalista

0.72

 Roskov

0.71

Erstellt

0.70

Personendaten

0.70

eqn

0.69

Tembelea

0.67

Activations Density 0.000%

tokens that consist of underscores or related patterns in data and formatting

Gravfile names and greek math notation

No Comments

No Known Activations