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

ãĥĨ

-0.71

ibr

-0.70

stream

-0.69

ood

-0.63

 Cornell

-0.63

['

-0.63

yi

-0.62

Â

-0.62

ãĥı

-0.62

ãĥ«

-0.60

POSITIVE LOGITS

 sacrific

0.84

 challeng

0.83

 comr

0.82

 contrace

0.82

insula

0.81

 comprom

0.80

 Palestin

0.78

 compe

0.77

 condem

0.77

worldly

0.76

Activations Density 0.000%

No Known Activations

This feature has no known activations.