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

lang

-0.75

OPLE

-0.74

byn

-0.71

utf

-0.69

boa

-0.68

bish

-0.66

poke

-0.66

gery

-0.66

coon

-0.65

lda

-0.65

POSITIVE LOGITS

 Innocent

0.86

ãĥ¼ãĥĨãĤ£

0.82

quished

0.79

 ãĤµãĥ¼ãĥĨãĤ£ãĥ¯ãĥ³

0.72

ãĥŁ

0.70

ãĤ³

0.68

æ©

0.67

ãĥ¯ãĥ³

0.67

çĶ°

0.66

 Tradable

0.65

Activations Density 0.000%

No Known Activations

This feature has no known activations.