INDEX

Explanations

words related to postal mail and fraud

oai_token-act-pair · gpt-3.5-turbo

references to mail-related activities or fraud

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

nen

-0.68

oulos

-0.64

 LORD

-0.64

 whiff

-0.62

ETF

-0.62

 Durham

-0.61

iets

-0.61

onomous

-0.61

aughs

-0.60

razil

-0.60

POSITIVE LOGITS

boxes

1.39

bag

1.19

 mail

1.16

bags

1.05

 mailbox

1.02

box

0.99

mail

0.95

Mail

0.87

email

0.86

 inbox

0.85

Activations Density 0.012%

words related to postal mail and fraud

references to mail-related activities or fraud

No Comments

No Known Activations