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

archment

-0.69

isode

-0.69

esville

-0.67

advertising

-0.67

RANT

-0.66

imaru

-0.66

TOP

-0.65

mington

-0.65

ihara

-0.65

warn

-0.65

POSITIVE LOGITS

0.70

ose

0.68

imble

0.62

 Reign

0.62

 accountability

0.60

 finite

0.59

âĹ¼

0.59

 Crescent

0.59

aird

0.59

 Lion

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.