INDEX

Explanations

references to physical holes or perforations

oai_token-act-pair · gpt-3.5-turbo

references to holes in various contexts, particularly in discussions about security and structure

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

 proposition

-0.88

 fundraiser

-0.83

 rematch

-0.81

ologue

-0.76

 declaration

-0.76

 standpoint

-0.75

 presidency

-0.74

 statement

-0.74

 blockade

-0.73

 competitor

-0.73

POSITIVE LOGITS

 holes

2.64

 loopholes

1.78

 gaps

1.69

holes

1.68

 openings

1.65

 cracks

1.55

 pockets

1.47

 clips

1.45

 spots

1.44

 layers

1.44

Activations Density 0.048%

references to physical holes or perforations

references to holes in various contexts, particularly in discussions about security and structure

No Comments

No Known Activations