INDEX

Explanations

references to public appearances or events

oai_token-act-pair · gpt-3.5-turbo

instances of the word "appearance" in various contexts

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

cise

-0.95

rals

-0.79

ced

-0.76

married

-0.73

EStreamFrame

-0.72

brace

-0.70

omb

-0.69

regate

-0.68

illed

-0.66

lease

-0.65

POSITIVE LOGITS

 appearance

1.35

 appearances

1.25

 Appearance

1.19

 flaws

0.82

lihood

0.79

 resemb

0.79

rul

0.78

DragonMagazine

0.77

 resemblance

0.77

 likeness

0.74

Activations Density 0.017%

references to public appearances or events

instances of the word "appearance" in various contexts

No Comments

No Known Activations