INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 Yose

-0.80

 Croat

-0.80

law

-0.64

 subcontract

-0.64

 quota

-0.63

 territ

-0.63

 swoop

-0.62

 Yemeni

-0.61

Zac

-0.61

 Burnett

-0.60

POSITIVE LOGITS

orce

0.76

isers

0.75

doi

0.75

 vulner

0.75

 Versus

0.73

adium

0.71

 Sorce

0.70

iding

0.69

é¾į

0.68

ided

0.68

Activations Density 0.000%

No Known Activations

This feature has no known activations.