INDEX

Explanations

mentions of locations or landmarks

oai_token-act-pair · gpt-3.5-turbo

references to specific locations or proper nouns

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

oks

-0.81

liter

-0.78

 Crypto

-0.77

BIT

-0.74

Rog

-0.74

kay

-0.73

 weed

-0.71

 Zoom

-0.71

VPN

-0.71

Merit

-0.70

POSITIVE LOGITS

 Pavilion

2.26

Pav

1.80

pav

1.64

ariat

1.32

 Southampton

1.30

ampton

1.22

 Hampton

1.21

 Portsmouth

1.21

 Fedora

1.12

 Pend

1.11

Activations Density 0.083%

mentions of locations or landmarks

references to specific locations or proper nouns

No Comments

No Known Activations