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

 oversight

-0.82

UGC

-0.75

OGR

-0.69

 prelim

-0.68

avez

-0.66

 chore

-0.64

 Nicaragua

-0.64

 BahÃ¡

-0.64

 appellate

-0.62

uala

-0.61

POSITIVE LOGITS

arin

0.74

mercial

0.74

stem

0.73

etry

0.69

wagon

0.67

arine

0.66

ritional

0.63

atively

0.62

izen

0.62

 teasp

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.