INDEX

Explanations

news-related and rumor-related terms and phrases

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate the occurrence of news reports or announcements

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

urally

-0.82

uits

-0.81

ĸļ

-0.78

unte

-0.78

uit

-0.77

emale

-0.75

untarily

-0.74

olon

-0.73

cientious

-0.73

erto

-0.73

POSITIVE LOGITS

 Inquis

1.03

 rumor

1.02

Rum

1.01

BBC

0.97

 gossip

0.97

 rumours

0.96

Sources

0.94

 rumors

0.93

 headlines

0.92

TMZ

0.92

Activations Density 0.795%

news-related and rumor-related terms and phrases

phrases that indicate the occurrence of news reports or announcements

No Comments

No Known Activations