INDEX

Explanations

prompts to 'watch' or 'visit website'

oai_token-act-pair · gpt-3.5-turbo

references to viewing or watching content, especially in media contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 lamb

-0.72

pled

-0.72

wards

-0.71

 perm

-0.67

etheless

-0.65

ework

-0.65

 whistle

-0.63

robe

-0.63

 downs

-0.61

sheets

-0.61

POSITIVE LOGITS

 Thumbnails

0.69

!--

0.69

 Kinnikuman

0.69

æ©

0.64

 Abandon

0.63

OTS

0.62

OSE

0.62

VK

0.62

tv

0.60

Brave

0.60

Activations Density 0.016%

prompts to 'watch' or 'visit website'

references to viewing or watching content, especially in media contexts

No Comments

No Known Activations