INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 atmosphere

-0.72

 unsett

-0.70

 amplified

-0.69

aqu

-0.68

ente

-0.68

ciplinary

-0.67

 outpost

-0.66

 overpower

-0.65

 aggress

-0.65

¬¼

-0.65

POSITIVE LOGITS

BSD

0.76

ĺħ

0.73

PsyNetMessage

0.70

fork

0.68

leigh

0.68

 Bulgar

0.67

ographically

0.67

 Veget

0.66

yip

0.66

gets

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.