INDEX

Explanations

the acronym "UL" with high levels of activation

oai_token-act-pair · gpt-3.5-turbo

abbreviations or acronyms associated with technical terminology or categories

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

 compass

-0.64

bro

-0.63

 cheek

-0.62

fam

-0.62

hip

-0.62

 circulation

-0.61

 lent

-0.61

Fam

-0.60

 apples

-0.60

 passer

-0.60

POSITIVE LOGITS

UL

4.33

ULE

2.33

ul

2.07

UL

1.67

ULAR

1.54

ULL

1.49

ULT

1.47

uls

1.46

ula

1.46

ulate

1.46

Activations Density 0.008%

the acronym "UL" with high levels of activation

abbreviations or acronyms associated with technical terminology or categories

No Comments

No Known Activations