INDEX

Explanations

phrases related to actions or directives

oai_token-act-pair · gpt-3.5-turbo

references to specific actions or decisions related to government and authority

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

ascript

-0.75

iste

-0.74

ipedia

-0.72

cies

-0.69

Redditor

-0.68

Åį

-0.68

icing

-0.67

cephal

-0.67

auga

-0.67

onom

-0.66

POSITIVE LOGITS

PER

1.82

GEN

1.82

 FIRST

1.82

DES

1.81

MIN

1.81

 STUD

1.80

WOR

1.79

MAN

1.79

THE

1.79

 TRANS

1.79

Activations Density 0.666%

phrases related to actions or directives

references to specific actions or decisions related to government and authority

No Comments

No Known Activations