INDEX

Explanations

keywords related to categorization or classification, such as "kinds" or "types"

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

isman

-0.61

Blade

-0.60

Leaks

-0.60

Trivia

-0.59

 Tycoon

-0.58

edia

-0.57

 Boards

-0.56

 appre

-0.56

PDATE

-0.55

 Commissioners

-0.55

POSITIVE LOGITS

etting

1.25

of

0.87

 thereof

0.81

paces

0.81

pace

0.80

hell

0.75

of

0.75

ot

0.71

classes

0.68

face

0.68

Activations Density 0.089%

keywords related to categorization or classification, such as "kinds" or "types"

No Comments

No Known Activations