INDEX

Explanations

references to a specific acronym "AU" associated with various contexts and entities

oai_token-act-pair · gpt-3.5-turbo

references to specific entities or organizations, particularly with the acronym "AU" and the term "AK."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

eches

-0.94

usable

-0.76

ishops

-0.75

WIND

-0.74

 horm

-0.73

mediately

-0.72

acter

-0.71

 Priebus

-0.70

ombs

-0.68

 Liver

-0.67

POSITIVE LOGITS

AK

2.23

AU

2.19

AU

1.99

arak

1.43

fal

1.31

cou

1.29

UA

1.05

 exotic

1.04

Kou

0.91

 ALEC

0.90

Activations Density 0.041%

references to a specific acronym "AU" associated with various contexts and entities

references to specific entities or organizations, particularly with the acronym "AU" and the term "AK."

No Comments

No Known Activations