INDEX

Explanations

proper nouns related to locations or buildings

oai_token-act-pair · gpt-3.5-turbo

references to specific plazas

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ERAL

-0.83

nesses

-0.75

liest

-0.73

gets

-0.72

nings

-0.67

manship

-0.67

%%%%

-0.67

loss

-0.65

ebook

-0.64

wcsstore

-0.64

POSITIVE LOGITS

 Plaza

1.33

 Hotel

1.00

 plaza

0.94

 Square

0.86

 Hilton

0.82

 Blvd

0.81

itect

0.78

 Corridor

0.78

atl

0.77

 Coliseum

0.75

Activations Density 0.011%

proper nouns related to locations or buildings

references to specific plazas

No Comments

No Known Activations