INDEX

Explanations

the word "bin"

oai_token-act-pair · gpt-3.5-turbo

references to Osama bin Laden

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

Tenn

-0.79

wright

-0.77

KY

-0.70

 Orleans

-0.67

 Grimes

-0.67

 Tennessee

-0.64

Phoenix

-0.62

 Slov

-0.62

Southern

-0.62

 Missouri

-0.62

POSITIVE LOGITS

bin

4.33

Bin

2.85

 bins

2.46

bin

1.75

ibn

1.51

 Osama

1.48

Ibn

1.27

Abu

0.93

 bucket

0.92

 Gaddafi

0.90

Activations Density 0.013%

the word "bin"

references to Osama bin Laden

No Comments

No Known Activations