INDEX

Explanations

informational phrases prompting for further reading or exploration

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate information seeking or inquiry

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

 Sabha

-0.75

 teasp

-0.71

Writer

-0.71

KK

-0.70

ÅĤ

-0.69

©¶æ¥µ

-0.67

mouth

-0.66

gam

-0.64

Creat

-0.64

rang

-0.63

POSITIVE LOGITS

how

0.74

ãĥ´ãĤ¡

0.70

 Yourself

0.67

 respecting

0.66

 Seller

0.65

outheast

0.64

 flavour

0.64

 whats

0.62

 yourself

0.60

 entitle

0.60

Activations Density 0.059%

informational phrases prompting for further reading or exploration

phrases that indicate information seeking or inquiry

No Comments

No Known Activations