INDEX

Explanations

phrases related to insider information

oai_token-act-pair · gpt-3.5-turbo

occurrences of an empty token, suggesting a focus on structural or formatting elements of the document

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

âĢ¢âĢ¢

-0.69

è¦ļéĨĴ

-0.68

ries

-0.66

erous

-0.66

llan

-0.66

 Baird

-0.65

RTX

-0.65

 Portuguese

-0.64

theless

-0.64

cha

-0.63

POSITIVE LOGITS

urance

1.41

omnia

1.34

ufficient

1.34

urrection

1.32

ights

1.24

iders

1.24

ulin

1.20

idious

1.16

ensitive

1.15

urances

1.11

Activations Density 0.023%

phrases related to insider information

occurrences of an empty token, suggesting a focus on structural or formatting elements of the document

No Comments

No Known Activations