INDEX

Explanations

phrases related to untrue or misleading content

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 @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

xual

-0.79

ands

-0.72

ires

-0.72

arching

-0.71

anded

-0.69

CONCLUS

-0.68

ained

-0.67

ean

-0.67

auts

-0.67

ENDED

-0.67

POSITIVE LOGITS

 fake

1.09

 Fake

0.98

fake

0.92

 phony

0.92

Fake

0.89

pas

0.85

 faked

0.84

 bait

0.80

 bogus

0.78

 clown

0.76

Activations Density 0.015%

phrases related to untrue or misleading content

references to fake news

No Comments

No Known Activations