INDEX

Explanations

mentions of negative or potentially harmful interactions between people

oai_token-act-pair · gpt-3.5-turbo

references to relationships and interactions with acquaintances

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

markets

-0.75

aer

-0.75

overe

-0.74

arde

-0.73

rahim

-0.72

rug

-0.72

ules

-0.71

YR

-0.69

enic

-0.68

adding

-0.68

POSITIVE LOGITS

 resemb

0.84

 resemblance

0.81

 familiarity

0.81

 acquaintance

0.80

 rapport

0.78

 glances

0.78

 communion

0.77

ance

0.75

DragonMagazine

0.74

 dealings

0.74

Activations Density 0.026%

mentions of negative or potentially harmful interactions between people

references to relationships and interactions with acquaintances

No Comments

No Known Activations