INDEX

Explanations

invalid or suspicious text entries such as email addresses

oai_token-act-pair · gpt-3.5-turbo

phrases indicating invalid email addresses

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

ILA

-0.74

krit

-0.72

renches

-0.70

zza

-0.69

hip

-0.68

ynthesis

-0.68

rooms

-0.66

ilic

-0.66

 bulldo

-0.66

toe

-0.65

POSITIVE LOGITS

 Invalid

0.93

 unfocusedRange

0.85

ities

0.79

 Thumbnails

0.78

ATED

0.75

 Email

0.72

LAND

0.72

AMI

0.70

Invalid

0.69

 Munich

0.67

Activations Density 0.016%

invalid or suspicious text entries such as email addresses

phrases indicating invalid email addresses

No Comments

No Known Activations