INDEX

Explanations

names or words containing initials followed by a numerical rank or position

oai_token-act-pair · gpt-3.5-turbo

abbreviations and initialisms within the text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

•EggList

No Comments

Negative Logits

inventoryQuantity

-0.79

 Contra

-0.77

soType

-0.74

 ASIC

-0.73

ENE

-0.73

 Fairfax

-0.72

 defic

-0.71

 antic

-0.69

 indisp

-0.68

 confir

-0.67

POSITIVE LOGITS

1.41

ks

1.39

1.30

kt

1.29

ked

1.26

king

1.24

KC

1.21

KS

1.21

KE

1.20

ke

1.20

Activations Density 0.242%

names or words containing initials followed by a numerical rank or position

abbreviations and initialisms within the text

No Comments

No Known Activations