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

haar

-0.90

mble

-0.86

ptoms

-0.84

llah

-0.84

schild

-0.84

gaard

-0.79

restricted

-0.78

beit

-0.76

reau

-0.74

burst

-0.74

POSITIVE LOGITS

 policy

1.17

 policies

1.00

 Policy

0.84

 Policies

0.77

policy

0.73

 Mayo

0.69

 Doodle

0.67

 Barbie

0.66

Ike

0.66

 Ellison

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.