INDEX

Explanations

words related to headlines and editorial content

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

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_23/width_131k/average_l0_101

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.23.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

帖最后由

-0.50

Var

-0.46

-0.45

PCV

-0.45

PSS

-0.44

((

-0.44

multer

-0.42

 comp

-0.41

Apakah

-0.41

Er

-0.40

POSITIVE LOGITS

 Headline

1.19

headline

1.15

Headline

1.11

 headline

1.09

 headlines

1.00

 Headlines

0.79

 getTitle

0.63

 eventName

0.62

 ModelExpression

0.62

 HEAD

0.60

Activations Density 0.005%

words related to headlines and editorial content

No Comments

No Known Activations