INDEX

Explanations

references to vague or unspecified subjects

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 13-gemmascope-res-65k

Configuration

google/gemma-scope-2b-pt-res/layer_13/width_65k/average_l0_75

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

65,536

Data Type

float32

Hook Name

blocks.13.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

UnusedPrivate

-0.85

 Arquivado

-0.83

Datuak

-0.81

xase

-0.74

 pleaſure

-0.73

titian

-0.72

 gales

-0.69

 uſe

-0.68

 Halls

-0.68

التالي

-0.68

POSITIVE LOGITS

 something

1.65

something

1.52

 Something

1.48

Something

1.48

ETHING

1.44

 SOMETHING

1.44

 else

1.28

 somethin

1.23

Somebody

1.20

 algo

1.02

Activations Density 0.029%

references to vague or unspecified subjects

No Comments

No Known Activations