INDEX

Explanations

word tokens related to user profiles

oai_token-act-pair · gpt-3.5-turbo

references to social media or public identities

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

EMS

-0.72

hers

-0.71

 Kingdoms

-0.68

Cla

-0.66

wic

-0.65

ieves

-0.65

Sew

-0.64

ccoli

-0.62

truth

-0.62

hens

-0.62

POSITIVE LOGITS

 profile

4.00

 profiles

3.05

profile

2.39

Profile

2.38

 Profile

2.37

 profiling

1.45

 biography

1.36

 footprint

1.25

 portrait

1.20

 persona

1.15

Activations Density 0.015%

word tokens related to user profiles

references to social media or public identities

No Comments

No Known Activations