INDEX

Explanations

words prefixed with "ex-", usually indicating a former or previous status

oai_token-act-pair · gpt-3.5-turbo

instances of the prefix "ex-" indicating former status or roles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 Sabha

-0.92

ãĤ¤ãĥĪ

-0.83

 GOODMAN

-0.83

 Flavoring

-0.79

GoldMagikarp

-0.74

jriwal

-0.73

arity

-0.70

WAYS

-0.69

hower

-0.68

OOD

-0.65

POSITIVE LOGITS

clamation

1.14

uber

1.09

ogenous

1.05

clud

0.99

orbit

0.99

oplan

0.99

terior

0.98

uding

0.95

uded

0.93

tern

0.92

Activations Density 0.009%

words prefixed with "ex-", usually indicating a former or previous status

instances of the prefix "ex-" indicating former status or roles

No Comments

No Known Activations