INDEX

Explanations

keyboard characters with patterns that encode a specific hidden message

oai_token-act-pair · gpt-3.5-turbo

phrases featuring numerical values or quantifications

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

roman

-0.67

PsyNetMessage

-0.65

ichick

-0.64

udeau

-0.64

iannopoulos

-0.63

 dissu

-0.61

ouble

-0.61

utical

-0.60

atis

-0.60

awei

-0.60

POSITIVE LOGITS

ÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂł

1.01

ÂłÂłÂłÂłÂłÂłÂłÂł

0.95

ÂłÂłÂłÂł

0.91

ÂłÂłÂł

0.89

Ear

0.80

ÂłÂł

0.76

Contents

0.71

Synopsis

0.66

ente

0.63

 Âł Âł

0.63

Activations Density 0.182%

keyboard characters with patterns that encode a specific hidden message

phrases featuring numerical values or quantifications

No Comments

No Known Activations