INDEX

Explanations

identification numbers and details related to official documentation

oai_token-act-pair · gpt-3.5-turbo

references to identification and personal information related to individuals

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Klux

-0.71

ipedia

-0.69

Louis

-0.68

folk

-0.67

video

-0.64

eah

-0.62

Yan

-0.61

NES

-0.61

nder

-0.61

ideos

-0.60

POSITIVE LOGITS

 each

1.37

 whichever

1.23

 your

0.95

 oneself

0.94

 whoever

0.91

 appropriate

0.89

 whatever

0.87

 applicable

0.86

 desired

0.85

any

0.85

Activations Density 0.579%

identification numbers and details related to official documentation

references to identification and personal information related to individuals

No Comments

No Known Activations