INDEX

Explanations

names of people, likely from news articles

oai_token-act-pair · gpt-3.5-turbo

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

tenance

-0.65

 WARN

-0.65

 hormonal

-0.63

 Terry

-0.62

 cereal

-0.62

Aid

-0.61

 Intermediate

-0.61

 Bland

-0.61

FR

-0.61

riber

-0.61

POSITIVE LOGITS

aret

2.27

aspers

1.96

MK

1.85

este

1.83

opard

1.80

alus

1.19

Latest

1.06

Kop

0.96

rero

0.96

 Latest

0.91

Activations Density 0.042%

names of people, likely from news articles

No Comments

No Known Activations