INDEX

Explanations

references to information that is considered false or deceptive

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 @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

-0.81

hens

-0.73

arching

-0.73

APTER

-0.71

azar

-0.71

}}}

-0.70

 Discuss

-0.68

ands

-0.68

ires

-0.67

Reviewed

-0.66

POSITIVE LOGITS

 fake

0.86

²¾

0.83

pas

0.83

 phony

0.72

 bait

0.71

ument

0.70

 Fake

0.70

 reef

0.70

ulously

0.67

eln

0.67

Activations Density 0.016%

references to information that is considered false or deceptive

references to "fake news."

No Comments

No Known Activations