INDEX

Explanations

phrases indicating a hypothetical scenario or condition

oai_token-act-pair · gpt-3.5-turbo

phrases that emphasize importance or significance related to subjects or actions

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

catentry

-0.66

 seams

-0.65

enhagen

-0.64

Comb

-0.63

 originate

-0.63

Orig

-0.62

 recoil

-0.61

 warrants

-0.60

ources

-0.60

 pockets

-0.59

POSITIVE LOGITS

 Yourself

0.80

 Patreon

0.67

soDeliveryDate

0.66

ivable

0.66

acons

0.66

readable

0.66

HER

0.65

 guiName

0.65

 yourselves

0.63

arer

0.62

Activations Density 0.131%

phrases indicating a hypothetical scenario or condition

phrases that emphasize importance or significance related to subjects or actions

No Comments

No Known Activations