INDEX

Explanations

repeated instances of the word "we"

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

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_20/width_131k/average_l0_114

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

-0.42

 foreign

-0.35

of

-0.33

ly

-0.33

 other

-0.33

 Autre

-0.33

덟

-0.30

 itself

-0.30

--

-0.30

 Foreign

-0.30

POSITIVE LOGITS

We

1.13

 CreateTagHelper

1.01

we

0.98

Mereka

0.96

We

0.93

 เรา

0.93

Мы

0.91

 ſind

0.90

 jsme

0.89

 ſei

0.89

Activations Density 0.185%

repeated instances of the word "we"

No Comments

No Known Activations