INDEX

Explanations

architectural features and details

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

Espèce

-0.63

roffene

-0.57

ocyan

-0.56

 Signalez

-0.54

 congen

-0.54

 Catho

-0.53

SNAPSHOT

-0.53

Mep

-0.52

 Jefus

-0.52

 unmarked

-0.50

POSITIVE LOGITS

 scroll

0.68

 NUKAT

0.63

 Winaray

0.60

 Scroll

0.59

 autorytatywna

0.59

setof

0.58

"}")

0.57

fromnode

0.56

служба

0.56

 StatelessWidget

0.54

Activations Density 0.020%

architectural features and details

No Comments

No Known Activations

architectural features and details

No Comments

No Known Activations