INDEX

Explanations

the word "rebut"

oai_token-act-pair · gpt-3.5-turbo

phrases related to the act of being removed or excluded

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

SHIP

-0.82

 Grayson

-0.65

 Carbuncle

-0.64

 CONTIN

-0.61

 redistributed

-0.61

 Jenna

-0.61

 attendance

-0.61

 Flowers

-0.59

 MacArthur

-0.59

 Barker

-0.59

POSITIVE LOGITS

ut

4.24

uts

2.48

UT

2.13

utor

1.95

uta

1.92

uti

1.88

utan

1.87

uto

1.84

uter

1.81

utt

1.79

Activations Density 0.011%

the word "rebut"

phrases related to the act of being removed or excluded

No Comments

No Known Activations