INDEX

Explanations

keywords related to titles or designations

oai_token-act-pair · gpt-3.5-turbo

mentions of the term "Attention."

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

 floating

-0.80

 frozen

-0.68

 crack

-0.68

 pale

-0.68

 Swiss

-0.67

 savings

-0.67

 stew

-0.66

 conscience

-0.66

 beans

-0.65

 slice

-0.64

POSITIVE LOGITS

Att

3.75

Att

1.94

att

1.80

ATT

1.71

Attach

1.52

ATT

1.44

 Attention

1.33

att

1.29

Attribute

1.27

Attack

1.27

Activations Density 0.017%

keywords related to titles or designations

mentions of the term "Attention."

No Comments

No Known Activations