INDEX

Explanations

spots or locations

oai_token-act-pair · gpt-3.5-turbo

repeated mentions of the word "spot."

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

 Means

-0.72

ibraries

-0.70

ORT

-0.67

ĸļ

-0.67

 Actions

-0.64

 Parties

-0.63

borg

-0.63

 Extensions

-0.62

ieri

-0.61

llo

-0.61

POSITIVE LOGITS

 spot

3.97

 spots

2.74

spot

2.50

 Spot

2.40

Spot

1.73

 berth

1.53

 spotting

1.49

 place

1.37

 slot

1.27

 location

1.20

Activations Density 0.008%

spots or locations

repeated mentions of the word "spot."

No Comments

No Known Activations