INDEX

Explanations

references to official or diplomatic buildings and objects

oai_token-act-pair · gpt-3.5-turbo

terms related to diplomatic or governmental contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

rama

-0.75

rand

-0.73

izoph

-0.70

corn

-0.69

 Lans

-0.68

wana

-0.68

reys

-0.67

ulic

-0.67

session

-0.67

oop

-0.66

POSITIVE LOGITS

folios

1.03

email

0.89

 consulate

0.84

mopolitan

0.79

âĸĳâĸĳ

0.79

ATIONS

0.78

xual

0.77

66666666

0.75

×Ĳ

0.74

ocity

0.73

Activations Density 0.029%

references to official or diplomatic buildings and objects

terms related to diplomatic or governmental contexts

No Comments

No Known Activations