INDEX

Explanations

words related to computer coding or language encoding

oai_token-act-pair · gpt-3.5-turbo

encoded characters or symbols that might indicate formatting or errors in text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

 Crosby

-0.82

 Pike

-0.79

 Kern

-0.74

 Jericho

-0.72

 Targ

-0.70

 Rothschild

-0.69

ovie

-0.68

 Phar

-0.68

 Azerbai

-0.67

 Trinidad

-0.67

POSITIVE LOGITS

¿

1.58

¹

1.48

¼

1.38

¤

1.34

°

1.33

ļ

1.25

Ī

1.25

¶

1.24

²

1.21

à¤

1.21

Activations Density 0.003%

words related to computer coding or language encoding

encoded characters or symbols that might indicate formatting or errors in text

No Comments

No Known Activations