INDEX

Explanations

phrases indicating continuation or urging to read more

oai_token-act-pair · gpt-3.5-turbo

prompts to continue reading or engaging with content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

*/(

-0.75

rams

-0.75

otom

-0.70

 formula

-0.68

bid

-0.66

 treat

-0.62

 rainy

-0.62

anas

-0.62

 makeshift

-0.60

 notoriously

-0.60

POSITIVE LOGITS

Continue

1.37

 Continued

0.91

 Continue

0.89

 Expand

0.81

00200000

0.77

Motion

0.77

 Progress

0.76

backer

0.74

Progress

0.74

 Continuing

0.74

Activations Density 0.008%

phrases indicating continuation or urging to read more

prompts to continue reading or engaging with content

No Comments

No Known Activations