INDEX

Explanations

phrases related to false information or deceit

oai_token-act-pair · gpt-3.5-turbo

references to fake news

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

xual

-0.95

hem

-0.75

night

-0.73

served

-0.73

riott

-0.71

onential

-0.69

onen

-0.66

pai

-0.65

 guiActiveUnfocused

-0.64

interrupted

-0.64

POSITIVE LOGITS

ument

1.02

 news

0.94

IDs

0.87

pas

0.87

 positives

0.81

ulent

0.79

 NEWS

0.73

 identities

0.70

ulence

0.70

outs

0.67

Activations Density 0.055%

phrases related to false information or deceit

references to fake news

No Comments

No Known Activations