INDEX

Explanations

headlines or titles containing strong and impactful words

oai_token-act-pair · gpt-3.5-turbo

proper nouns and names

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

 Franks

-0.69

Maker

-0.68

icans

-0.62

naires

-0.61

fox

-0.61

 makers

-0.60

 Pike

-0.60

 ponies

-0.60

 Maker

-0.58

 coats

-0.58

POSITIVE LOGITS

nown

0.89

entimes

0.86

clair

0.85

withstanding

0.84

¥ŀ

0.83

stantial

0.76

rely

0.75

specified

0.73

lihood

0.72

ntil

0.71

Activations Density 0.288%

headlines or titles containing strong and impactful words

proper nouns and names

No Comments

No Known Activations