INDEX

Explanations

words related to secretive or hidden actions or intentions

oai_token-act-pair · gpt-3.5-turbo

references to lurking or hidden elements, particularly in a suspenseful or dramatic context

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ãĥ¼ãĥ³

-0.75

atibility

-0.74

elson

-0.72

 Luther

-0.70

 Errors

-0.69

ELS

-0.67

 Fathers

-0.66

 Patriarch

-0.63

 Pacers

-0.62

arov

-0.62

POSITIVE LOGITS

lur

0.95

uten

0.80

iously

0.78

 lurking

0.76

kers

0.75

ormal

0.73

itas

0.71

 Jagu

0.69

cher

0.69

der

0.68

Activations Density 0.012%

words related to secretive or hidden actions or intentions

references to lurking or hidden elements, particularly in a suspenseful or dramatic context

No Comments

No Known Activations