INDEX

Explanations

references to entities or places with the word "Royal" in them

oai_token-act-pair · gpt-3.5-turbo

mentions of "Royal" in various contexts

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

 Hispanics

-0.81

WARE

-0.74

 Latinos

-0.74

stand

-0.71

ide

-0.71

gger

-0.70

src

-0.69

heit

-0.67

nyder

-0.66

 Newt

-0.66

POSITIVE LOGITS

 Royal

3.74

Royal

3.05

 royal

2.05

 Queen

1.37

roy

1.35

 Royale

1.35

 Majesty

1.35

RAF

1.33

 Princess

1.29

 Imperial

1.28

Activations Density 0.010%

references to entities or places with the word "Royal" in them

mentions of "Royal" in various contexts

No Comments

No Known Activations