INDEX

Explanations

religious/spiritual-related terms, specifically related to prayers

oai_token-act-pair · gpt-3.5-turbo

references to prayers and expressions of support or sympathy

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

ership

-0.74

culosis

-0.70

cer

-0.69

stood

-0.68

mental

-0.67

stadt

-0.66

access

-0.65

onomy

-0.65

cele

-0.64

transfer

-0.64

POSITIVE LOGITS

 prayers

1.25

mith

0.92

creen

0.87

peed

0.86

terday

0.82

chool

0.78

aurus

0.77

 redes

0.77

ugi

0.76

uits

0.75

Activations Density 0.019%

religious/spiritual-related terms, specifically related to prayers

references to prayers and expressions of support or sympathy

No Comments

No Known Activations