INDEX

Explanations

locations, specifically focusing on cities

oai_token-act-pair · gpt-3.5-turbo

references to locations, specifically cities in Brazil and China

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

UC

-0.79

nt

-0.76

Lear

-0.76

apse

-0.73

Kyle

-0.69

APS

-0.69

TIT

-0.69

NAS

-0.68

UM

-0.67

REE

-0.67

POSITIVE LOGITS

 Janeiro

1.84

 Lumpur

1.24

ascus

1.08

 Aires

0.94

 Paulo

0.86

 Jinping

0.82

raltar

0.81

 Pradesh

0.81

ĳå£«

0.81

berra

0.78

Activations Density 0.020%

locations, specifically focusing on cities

references to locations, specifically cities in Brazil and China

No Comments

No Known Activations