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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

cis

-0.76

 Kats

-0.70

 Dragons

-0.68

 Chap

-0.65

 sands

-0.65

 Assy

-0.64

 Armen

-0.64

 unden

-0.64

Redditor

-0.64

angan

-0.63

POSITIVE LOGITS

WARD

0.80

otiation

0.80

*/(

0.74

eering

0.70

agra

0.69

liness

0.69

alloween

0.69

haul

0.68

EStream

0.68

MRI

0.68

Activations Density 0.000%

No Known Activations

This feature has no known activations.