INDEX

Explanations

references to religious or sacred buildings, specifically temples

oai_token-act-pair · gpt-3.5-turbo

mentions of temples

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

BAT

-0.79

-0.77

CAP

-0.73

JUST

-0.73

ACP

-0.72

Published

-0.72

OVER

-0.70

SPI

-0.70

ANA

-0.69

DOS

-0.69

POSITIVE LOGITS

 temple

1.51

 temples

1.37

 maiden

0.90

 shrine

0.88

 deity

0.85

 altar

0.84

 courtyard

0.81

esan

0.80

 ceremonies

0.78

 worshipped

0.78

Activations Density 0.004%

references to religious or sacred buildings, specifically temples

mentions of temples

No Comments

No Known Activations