INDEX

Explanations

phrases related to technical specifications or descriptions of products or items, often including measurements or features

oai_token-act-pair · gpt-3.5-turbo

numerical measurements and associated descriptors

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

appropriate

-0.65

rather

-0.63

Ì¶

-0.63

`.

-0.62

ilaterally

-0.61

serious

-0.60

verified

-0.60

agree

-0.60

appropriately

-0.59

adequ

-0.58

POSITIVE LOGITS

 consists

1.36

is

1.29

 comprises

1.22

 differs

1.19

 tends

1.17

 represents

1.17

 includes

1.16

isn

1.16

has

1.15

 resembles

1.15

Activations Density 0.516%

phrases related to technical specifications or descriptions of products or items, often including measurements or features

numerical measurements and associated descriptors

No Comments

No Known Activations