INDEX

Explanations

the word "email"

oai_token-act-pair · gpt-3.5-turbo

mentions of "email" and related terms concerning subscriptions

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

TRY

-0.75

Yen

-0.71

pu

-0.67

 Templar

-0.67

liga

-0.67

stru

-0.65

 Mercer

-0.63

hma

-0.63

 Seraph

-0.62

itute

-0.61

POSITIVE LOGITS

email

0.94

 inbox

0.94

 address

0.87

antha

0.81

 email

0.80

 notification

0.79

 notifications

0.77

 addresses

0.77

 alerts

0.77

 mail

0.76

Activations Density 0.016%

the word "email"

mentions of "email" and related terms concerning subscriptions

No Comments

No Known Activations