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

 confir

-0.86

>)

-0.75

inished

-0.74

>]

-0.73

ngth

-0.73

uters

-0.73

urry

-0.71

 facult

-0.70

ONT

-0.70

aunder

-0.70

POSITIVE LOGITS

 Lama

0.79

 Lomb

0.75

 Judd

0.71

 Dawn

0.71

 Mari

0.69

Jol

0.68

stage

0.68

Äĩ

0.68

vati

0.67

Vas

0.65

Activations Density 0.000%

No Known Activations

This feature has no known activations.