INDEX

Explanations

references to specific individuals or entities, potentially related to news articles

oai_token-act-pair · gpt-3.5-turbo

instances of the abbreviation "CH" followed by numeric values

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

berman

-0.96

ãĤ´ãĥ³

-0.79

ãĥ¼ãĥĨãĤ£

-0.76

 Lithuan

-0.65

 autobi

-0.64

 centrif

-0.64

DragonMagazine

-0.63

gaard

-0.62

 stocking

-0.61

Äĩ

-0.60

POSITIVE LOGITS

APTER

1.06

ERY

0.96

CH

0.96

ECK

0.96

AIN

0.95

ANGE

0.93

AIR

0.93

ASE

0.92

AMP

0.92

OOL

0.89

Activations Density 0.006%

references to specific individuals or entities, potentially related to news articles

instances of the abbreviation "CH" followed by numeric values

No Comments

No Known Activations