INDEX

Explanations

phrases related to political and social scandals

oai_token-act-pair · gpt-3.5-turbo

references to news reports and journalistic investigations

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

•GPT2 L10 resid: 'LRRRRRLLRRRL'[:-2]

•GPT2 L10 resid: 'LRRRRRLLRRRL'[:-3]

No Comments

Negative Logits

 Expression

-0.78

anguages

-0.78

casters

-0.77

 learners

-0.75

anguage

-0.74

ignty

-0.72

aer

-0.69

gans

-0.68

 Languages

-0.67

unicip

-0.67

POSITIVE LOGITS

 scandal

1.63

 scandals

1.48

 saga

1.47

 affair

1.43

 intrigue

1.42

 tale

1.38

 allegations

1.37

 storyline

1.31

 revelations

1.31

 story

1.29

Activations Density 0.756%

phrases related to political and social scandals

references to news reports and journalistic investigations

No Comments

No Known Activations