INDEX

Explanations

fraud-related terms and concepts

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

terms related to various forms of fraud and misconduct

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 12-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.hook_resid_post

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.11.hook_resid_post

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

leigh

-0.65

 childbirth

-0.65

 bruising

-0.65

Dak

-0.64

 vain

-0.61

 Ange

-0.60

answer

-0.59

âĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢ

-0.59

 tranqu

-0.57

Alpha

-0.56

POSITIVE LOGITS

ulent

1.42

ulence

1.37

sters

1.23

ster

1.05

ul

1.03

 perpetrated

1.03

 scandals

1.00

 allegations

0.99

 schemes

0.94

 charges

0.93

Activations Density 0.101%

fraud-related terms and concepts

terms related to various forms of fraud and misconduct

No Comments

No Known Activations