INDEX

Explanations

discussion of urgent or important topics that are "must watch."

oai_token-act-pair · gpt-3.5-turbo

references to "watch" or "must watch" content in a document

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 pawn

-0.73

 lawn

-0.69

 nowhere

-0.69

 hust

-0.68

 crushing

-0.67

 shred

-0.66

 stomp

-0.65

 carpet

-0.65

 alienated

-0.65

 recl

-0.65

POSITIVE LOGITS

 WATCH

1.57

 Watching

1.04

 Thumbnails

0.99

 WATCHED

0.96

LIST

0.95

 VIDEOS

0.93

 Watch

0.91

WATCH

0.90

 Related

0.88

 IMAGES

0.86

Activations Density 0.006%

discussion of urgent or important topics that are "must watch."

references to "watch" or "must watch" content in a document

No Comments

No Known Activations