INDEX

Explanations

words related to things that are very highly emphasized or intensely focused on

oai_token-act-pair · gpt-3.5-turbo

intensifiers related to popularity or significance

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

ividual

-0.73

verts

-0.71

igi

-0.69

spection

-0.69

 Therapy

-0.68

 Farmers

-0.66

 VIDEOS

-0.66

 Dwell

-0.66

 Thumbnails

-0.65

ylene

-0.65

POSITIVE LOGITS

bitious

0.80

 profitable

0.79

 popular

0.75

 improbable

0.73

efully

0.73

 underrated

0.73

 unlikely

0.72

 impractical

0.72

 ambitious

0.72

 Important

0.71

Activations Density 0.019%

words related to things that are very highly emphasized or intensely focused on

intensifiers related to popularity or significance

No Comments

No Known Activations