INDEX

Explanations

words related to fraud

oai_token-act-pair · gpt-3.5-turbo

references to various forms of fraud

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 Ange

-0.74

âĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢ

-0.69

ŃĶ

-0.63

 Quartz

-0.62

cki

-0.59

 Afric

-0.59

 Aram

-0.59

 Borders

-0.58

 sunset

-0.58

 chimpanzees

-0.58

POSITIVE LOGITS

ulent

1.67

ulence

1.56

sters

1.44

ster

1.32

ul

1.15

ulus

1.06

uli

1.02

raud

1.02

 perpetrated

1.02

ulators

0.98

Activations Density 0.046%

words related to fraud

references to various forms of fraud

No Comments

No Known Activations