INDEX

Explanations

declarative sentences discussing responsibility and admitting wrongdoing

oai_token-act-pair · gpt-3.5-turbo

phrases related to the consequences of wrongdoing and the process of acknowledging mistakes

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

Not in Any Lists

No Comments

Negative Logits

 Emails

-0.76

 Folder

-0.72

anmar

-0.70

 folders

-0.70

taboola

-0.69

PokÃ©

-0.67

 Sprint

-0.66

ursday

-0.66

 Boom

-0.65

PCB

-0.64

POSITIVE LOGITS

 morally

1.15

 thereby

1.06

 acquies

1.03

 oneself

1.02

 intolerable

0.99

 impover

0.95

 coerc

0.94

 emanc

0.94

 sufficiently

0.93

iberal

0.92

Activations Density 0.709%

declarative sentences discussing responsibility and admitting wrongdoing

phrases related to the consequences of wrongdoing and the process of acknowledging mistakes

No Comments

No Known Activations