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

rongh

-0.73

ioxid

-0.72

erves

-0.70

ãĤ¤ãĥĪ

-0.68

abouts

-0.66

ablishment

-0.66

 Rosenthal

-0.65

 Moss

-0.65

IRA

-0.62

 estates

-0.62

POSITIVE LOGITS

ama

0.66

urnal

0.64

maid

0.63

toe

0.62

lav

0.62

unction

0.62

aid

0.62

isively

0.61

adic

0.61

Äĩ

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.