INDEX

Explanations

references to pairs or sets of things that are similar or related

oai_token-act-pair · gpt-3.5-turbo

references to twins

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

UME

-0.78

anwhile

-0.73

CoC

-0.72

WI

-0.70

andise

-0.68

arcity

-0.68

ugu

-0.66

 Aware

-0.65

 Clin

-0.64

 Explain

-0.64

POSITIVE LOGITS

ning

0.96

ned

0.94

 twin

0.90

 brother

0.85

 brothers

0.80

 towers

0.76

 twins

0.75

 sister

0.75

 fold

0.74

hered

0.70

Activations Density 0.008%

references to pairs or sets of things that are similar or related

references to twins

No Comments

No Known Activations