INDEX

Explanations

references to titles or names indicating authority or nobility

oai_token-act-pair · gpt-3.5-turbo

references to titles or positions of authority, specifically the term "Lord."

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

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

ÄŁ

-0.78

SAP

-0.70

 fingerprint

-0.68

 capac

-0.65

thumbnails

-0.65

edin

-0.65

 reciproc

-0.64

Cub

-0.64

 multip

-0.63

fg

-0.63

POSITIVE LOGITS

 Lord

3.85

Lord

3.06

 LORD

2.55

 Lords

2.38

 lord

2.30

 Lady

1.73

 lords

1.72

 Savior

1.66

Lady

1.60

lord

1.49

Activations Density 0.014%

references to titles or names indicating authority or nobility

references to titles or positions of authority, specifically the term "Lord."

No Comments

No Known Activations