INDEX

Explanations

mentions of royal titles and figures, particularly those related to princes

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

 Exper

-0.85

 Nationwide

-0.85

 Transportation

-0.84

 Albuquerque

-0.83

 Terminal

-0.83

lihood

-0.79

 Neon

-0.79

 Spotlight

-0.79

 Supp

-0.78

 Project

-0.78

POSITIVE LOGITS

 prince

2.86

 princess

2.68

 princes

2.32

 lords

2.11

 lord

2.06

 knight

1.98

 knights

1.88

 nobles

1.85

 castles

1.84

 nobility

1.81

Activations Density 0.040%

mentions of royal titles and figures, particularly those related to princes

No Comments

No Known Activations