INDEX

Explanations

phrases related to political events or places

oai_token-act-pair · gpt-3.5-turbo

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

BALL

-0.70

wagon

-0.67

DEV

-0.67

 WARN

-0.64

Plex

-0.62

Cass

-0.61

log

-0.61

spin

-0.60

Plugin

-0.59

yne

-0.58

POSITIVE LOGITS

 Hong

3.78

Hong

3.47

 Guang

2.22

HK

2.10

 Shanghai

1.91

 Singapore

1.91

 Taiwan

1.82

 Beijing

1.78

 Tian

1.77

 Taiwanese

1.69

Activations Density 0.019%

phrases related to political events or places

No Comments

No Known Activations