INDEX

Explanations

phrases related to information or news reports

oai_token-act-pair · gpt-3.5-turbo

consistent mentions of 'reports'

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

 Klux

-0.72

 wedge

-0.71

chest

-0.67

 palate

-0.66

othing

-0.66

 ILCS

-0.65

ococ

-0.64

abies

-0.63

vil

-0.63

asus

-0.62

POSITIVE LOGITS

 reports

1.14

reports

1.03

 Reports

0.88

report

0.85

Reports

0.85

udo

0.82

 report

0.82

books

0.77

uggest

0.76

Reporting

0.74

Activations Density 0.026%

phrases related to information or news reports

consistent mentions of 'reports'

No Comments

No Known Activations