INDEX

Explanations

words or phrases associated with invalid emails

oai_token-act-pair · gpt-3.5-turbo

references to controversial or criminal activities

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

 favoring

-0.86

 recognizing

-0.86

cius

-0.78

 favors

-0.78

 calibrated

-0.77

iations

-0.76

 normalized

-0.76

 constituted

-0.75

 stabilize

-0.75

 recognized

-0.75

POSITIVE LOGITS

Liverpool

1.01

 Notting

1.01

 Ukip

1.00

 Newcastle

0.98

DUP

0.98

 Nicola

0.95

Tube

0.95

Labour

0.95

GET

0.93

Scotland

0.92

Activations Density 0.217%

words or phrases associated with invalid emails

references to controversial or criminal activities

No Comments

No Known Activations