INDEX

Explanations

directives starting with the word "Please."

oai_token-act-pair · gpt-3.5-turbo

commands and requests for action

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

ailability

-0.76

EStreamFrame

-0.76

azeera

-0.75

ilogy

-0.73

Ö¼

-0.73

ufact

-0.71

pas

-0.69

aturated

-0.65

erial

-0.65

capacity

-0.65

POSITIVE LOGITS

sir

0.83

 PLEASE

0.68

iquette

0.67

 politely

0.66

estine

0.65

":[{"

0.65

 Submit

0.65

 ORDER

0.64

!!!!

0.64

 acknowled

0.64

Activations Density 0.080%

directives starting with the word "Please."

commands and requests for action

No Comments

No Known Activations