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

ioxide

-0.84

igslist

-0.81

anooga

-0.79

£ı

-0.76

ileaks

-0.75

ideo

-0.75

AMI

-0.74

etting

-0.73

merce

-0.73

POSITIVE LOGITS

 Rend

0.70

mann

0.67

 Numbers

0.63

 delegation

0.63

 Nazis

0.62

 cones

0.61

Hen

0.58

number

0.57

 dise

0.56

 nutshell

0.56

Activations Density 0.000%

No Known Activations

This feature has no known activations.