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

umenthal

-0.83

worms

-0.75

itudinal

-0.70

cffff

-0.70

orry

-0.67

enture

-0.67

 pains

-0.67

ochond

-0.66

æ©Ł

-0.66

webkit

-0.66

POSITIVE LOGITS

 reperto

0.65

Zak

0.64

Nun

0.63

 ordinances

0.62

 services

0.60

 Lara

0.60

 ministries

0.60

 communities

0.59

services

0.59

 Shap

0.59

Activations Density 0.000%

No Known Activations

This feature has no known activations.