INDEX

Explanations

phrases related to being inspired or motivated

oai_token-act-pair · gpt-3.5-turbo

references to law enforcement individuals or titles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 Nanto

-0.81

bos

-0.80

days

-0.78

DAY

-0.76

DEN

-0.74

town

-0.73

çĦ

-0.70

child

-0.68

\\\\\\\\

-0.67

SHIP

-0.67

POSITIVE LOGITS

iration

1.57

 Insp

1.41

Insp

1.18

ired

0.96

isitions

0.87

iring

0.87

iliate

0.86

iasco

0.86

iral

0.85

anmar

0.84

Activations Density 0.005%

phrases related to being inspired or motivated

references to law enforcement individuals or titles

No Comments

No Known Activations