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

 burner

-0.73

 Birch

-0.68

 Inher

-0.67

issue

-0.66

hay

-0.65

 Sodium

-0.64

 commons

-0.64

ensitive

-0.64

 fray

-0.64

CTR

-0.63

POSITIVE LOGITS

TAIN

0.94

SU

0.83

²¾

0.79

å§«

0.71

borgh

0.69

ulously

0.68

essel

0.67

 mosqu

0.66

ãĥ´

0.65

nant

0.65

Activations Density 0.000%

No Known Activations

This feature has no known activations.