INDEX

Explanations

No Explanations Found

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

ļéĨĴ

-0.83

lished

-0.83

adobe

-0.78

amaz

-0.77

iets

-0.76

ãĤ¼ãĤ¦ãĤ¹

-0.74

zers

-0.72

 caut

-0.72

ACP

-0.72

ãħĭ

-0.72

POSITIVE LOGITS

 syndrome

0.70

("

0.68

 immediately

0.63

 associated

0.63

 intended

0.62

 sideline

0.62

 touchscreen

0.60

 allegedly

0.59

 effect

0.59

 infield

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.