INDEX

Explanations

mentions of specific camps, likely related to refugees or detainment

oai_token-act-pair · gpt-3.5-turbo

references to various types of camps, particularly refugee and detention camps

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

 Ruler

-0.71

ibe

-0.68

 reprint

-0.62

ķ

-0.61

 RELEASE

-0.60

mbol

-0.59

 Shape

-0.59

mary

-0.57

 aver

-0.57

phone

-0.57

POSITIVE LOGITS

 camps

4.11

 camp

2.84

camp

2.00

 Camp

1.90

 encamp

1.85

 tents

1.76

Camp

1.75

 camping

1.52

 tent

1.21

 trenches

1.17

Activations Density 0.013%

mentions of specific camps, likely related to refugees or detainment

references to various types of camps, particularly refugee and detention camps

No Comments

No Known Activations