INDEX

Explanations

phrases indicating progression or advancement

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

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_33/width_131k/average_l0_91

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.33.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 kasarigan

-0.62

Portály

-0.52

पया

-0.47

最快更新

-0.46

BufferException

-0.44

 viewDidLoad

-0.43

 المعيارى

-0.43

ConstraintMaker

-0.41

 noDo

-0.40

gunos

-0.40

POSITIVE LOGITS

 level

1.59

level

1.25

 levels

1.16

 LEVEL

1.02

levels

1.00

 livello

0.98

 niveau

0.96

Level

0.94

レベル

0.93

 nivel

0.92

Activations Density 0.011%

phrases indicating progression or advancement

No Comments

No Known Activations