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

binding

-0.77

 Icelandic

-0.72

bah

-0.71

 locality

-0.68

hook

-0.68

jit

-0.67

intosh

-0.67

motion

-0.66

sein

-0.66

asta

-0.65

POSITIVE LOGITS

clerosis

0.73

orest

0.71

 Rounds

0.70

oice

0.67

ervatives

0.65

 Percent

0.64

 Include

0.64

Family

0.63

clude

0.63

clud

0.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.