INDEX

Explanations

phrases indicating unnecessary but emphasized actions or situations

oai_token-act-pair · gpt-3.5-turbo

phrases indicating redundancy or unnecessary information

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

backed

-0.73

locking

-0.70

ingle

-0.67

ank

-0.66

oving

-0.65

miah

-0.65

 mailbox

-0.62

wana

-0.62

rams

-0.61

ammed

-0.60

POSITIVE LOGITS

fits

0.79

 Canaver

0.77

Osw

0.77

vity

0.76

vation

0.72

fer

0.70

appropriately

0.69

iously

0.68

 modesty

0.68

erest

0.67

Activations Density 0.049%

phrases indicating unnecessary but emphasized actions or situations

phrases indicating redundancy or unnecessary information

No Comments

No Known Activations