INDEX

Explanations

the word "ief" with a preference for higher activations

oai_token-act-pair · gpt-3.5-turbo

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

asket

-0.70

 lifetime

-0.63

umm

-0.61

Downloadha

-0.60

 compelling

-0.60

 provocative

-0.58

 physical

-0.58

 repetitive

-0.57

 rotating

-0.57

present

-0.57

POSITIVE LOGITS

ief

1.46

doms

1.21

ornia

1.05

iciary

0.98

llor

0.88

iev

0.86

lings

0.83

uin

0.82

ortun

0.81

gey

0.78

Activations Density 0.005%

the word "ief" with a preference for higher activations

No Comments

No Known Activations