INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

ensibly

-0.80

ablishment

-0.70

utton

-0.67

cape

-0.65

urated

-0.64

akable

-0.64

stand

-0.64

atel

-0.64

 refres

-0.63

ished

-0.63

POSITIVE LOGITS

products

0.67

 Household

0.67

emouth

0.66

soType

0.64

 Fifth

0.63

 Lamar

0.63

 equivalents

0.62

 vans

0.62

Street

0.61

 persuaded

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.