INDEX

Explanations

references to television channels

oai_token-act-pair · gpt-3.5-turbo

references to various television channels

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

 Raiders

-0.83

Pri

-0.63

 Hockey

-0.63

 hers

-0.62

 Giants

-0.60

 Janeiro

-0.59

 Marg

-0.59

 Bruins

-0.58

 Raider

-0.57

engu

-0.57

POSITIVE LOGITS

 channels

1.38

 channel

1.01

icut

0.92

lain

0.87

annels

0.87

channel

0.84

izoph

0.83

iew

0.82

icate

0.79

Channel

0.79

Activations Density 0.007%

references to television channels

references to various television channels

No Comments

No Known Activations