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

 chall

-0.74

 tremend

-0.74

 proble

-0.69

 advances

-0.67

 coord

-0.66

 climbs

-0.65

uers

-0.64

 srfAttach

-0.63

fri

-0.63

 leaps

-0.63

POSITIVE LOGITS

Ware

0.68

lance

0.66

gob

0.65

ewater

0.64

waukee

0.63

ItemImage

0.62

 Jaguar

0.62

 Luxem

0.61

 Merchants

0.61

ansk

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.