INDEX

Explanations

geographical locations, with a focus on Hong Kong

oai_token-act-pair · gpt-3.5-turbo

mentions of Hong Kong

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 inference

-0.66

 enhancement

-0.64

 instr

-0.63

 blow

-0.63

Ø§ÙĦ

-0.62

 marker

-0.61

nesota

-0.61

 wedge

-0.61

 nause

-0.60

 mileage

-0.60

POSITIVE LOGITS

 Kong

1.57

awei

1.04

zhou

0.99

moon

0.91

nam

0.90

jin

0.89

wu

0.85

wei

0.85

 Hong

0.83

ju

0.83

Activations Density 0.013%

geographical locations, with a focus on Hong Kong

mentions of Hong Kong

No Comments

No Known Activations