INDEX

Explanations

mentions of talk or interview shows along with specific details like names of the shows or hosts

oai_token-act-pair · gpt-3.5-turbo

references to television shows and appearances

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

uilt

-0.76

auga

-0.76

 stagn

-0.69

oras

-0.69

uay

-0.68

 labou

-0.68

orius

-0.68

 defective

-0.68

 leve

-0.67

ritis

-0.66

POSITIVE LOGITS

 aired

1.37

 airs

1.31

 television

1.23

 broadcasts

1.23

TV

1.23

 airing

1.22

ABC

1.22

NBC

1.20

NBC

1.19

levision

1.18

Activations Density 0.523%

mentions of talk or interview shows along with specific details like names of the shows or hosts

references to television shows and appearances

No Comments

No Known Activations