INDEX

Explanations

phrases related to printing or publication

oai_token-act-pair · gpt-3.5-turbo

terms associated with print media and print-related actions

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

ogun

-0.76

akra

-0.74

«ĺ

-0.72

arya

-0.71

ovi

-0.70

omez

-0.69

itimate

-0.68

radius

-0.67

blance

-0.65

ĪĴ

-0.65

POSITIVE LOGITS

DragonMagazine

1.12

é¾įå

0.73

 Copy

0.71

Plex

0.70

 Unloaded

0.68

âĢ¢âĢ¢

0.66

ilts

0.66

DX

0.64

ishable

0.64

 linen

0.64

Activations Density 0.284%

phrases related to printing or publication

terms associated with print media and print-related actions

No Comments

No Known Activations