INDEX

Explanations

instances of the abbreviation "DE" followed by a number indicating intensity or importance

oai_token-act-pair · gpt-3.5-turbo

the abbreviation "DE" in various contexts

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

HE

-0.85

WithNo

-0.74

hots

-0.67

 Lakers

-0.67

gor

-0.66

Bird

-0.66

socket

-0.66

Bos

-0.65

lining

-0.65

HE

-0.65

POSITIVE LOGITS

DE

2.11

 Julie

1.34

 Charlottesville

1.27

DE

1.17

 Ricardo

1.12

DEA

1.11

 Concord

1.02

 Gazette

1.01

 Hydra

0.90

ABE

0.83

Activations Density 0.028%

instances of the abbreviation "DE" followed by a number indicating intensity or importance

the abbreviation "DE" in various contexts

No Comments

No Known Activations