INDEX

Explanations

references to royal or regal imagery such as thrones and palaces

oai_token-act-pair · gpt-3.5-turbo

terms related to thrones and royal authority

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

ver

-0.71

professional

-0.65

 Reported

-0.64

Kuh

-0.63

Ver

-0.63

 Exper

-0.62

reports

-0.61

nell

-0.61

 Local

-0.61

Opt

-0.60

POSITIVE LOGITS

 throne

4.06

 Throne

2.51

 crown

1.55

 palace

1.51

 monarchy

1.35

 dynasty

1.24

 kings

1.23

 king

1.23

 mantle

1.21

 kingdom

1.21

Activations Density 0.016%

references to royal or regal imagery such as thrones and palaces

terms related to thrones and royal authority

No Comments

No Known Activations