INDEX

Explanations

phrases related to news captions

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

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

 unsett

-0.79

 concussion

-0.73

etheless

-0.73

 knees

-0.71

 peanuts

-0.70

warm

-0.70

 rejuven

-0.70

metic

-0.69

bats

-0.69

shake

-0.68

POSITIVE LOGITS

 Courtesy

1.44

Courtesy

1.01

 Shutterstock

1.00

 COURT

0.99

 Image

0.98

 REUTERS

0.96

 Provided

0.94

 Photograph

0.92

 toggle

0.91

 Â©

0.90

Activations Density 0.008%

phrases related to news captions

No Comments

No Known Activations