INDEX

Explanations

words related to negations and skepticism

oai_token-act-pair · gpt-3.5-turbo

occurrences of the prefix "imper" suggesting notions of permanence or denial

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

anwhile

-0.77

gdala

-0.75

Ou

-0.68

 Hath

-0.66

enne

-0.64

wagen

-0.63

hare

-0.62

 Stand

-0.62

esis

-0.60

 Carbuncle

-0.59

POSITIVE LOGITS

manent

1.36

vious

1.34

missible

1.32

ishable

1.20

bably

1.01

mental

0.99

ceivable

0.98

genic

0.94

cept

0.92

miss

0.90

Activations Density 0.032%

words related to negations and skepticism

occurrences of the prefix "imper" suggesting notions of permanence or denial

No Comments

No Known Activations