INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

ceans

-0.69

hester

-0.68

vest

-0.67

 corrid

-0.66

GY

-0.66

 Aval

-0.66

vez

-0.64

liga

-0.63

âĢİ

-0.62

vel

-0.61

POSITIVE LOGITS

 Sons

0.74

andum

0.73

ENC

0.70

 wrongful

0.68

 divest

0.67

DPR

0.66

conn

0.66

 succession

0.64

WS

0.63

 regards

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.