INDEX

Explanations

phrases relating to considering, evaluating, or giving opinions on various topics

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

ICLE

-0.68

eer

-0.68

 Gutenberg

-0.68

nered

-0.66

 Broken

-0.63

BLE

-0.61

 Brist

-0.61

chief

-0.60

 Kart

-0.60

etheless

-0.60

POSITIVE LOGITS

 weigh

1.10

 weighing

0.89

 weights

0.84

ighed

0.84

 heaviest

0.82

 weighed

0.78

 heavily

0.77

 down

0.75

 heavier

0.74

enance

0.73

Activations Density 0.044%

phrases relating to considering, evaluating, or giving opinions on various topics

No Comments

No Known Activations