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

itely

-0.70

 Lobby

-0.69

bub

-0.66

HUD

-0.62

noon

-0.61

fo

-0.61

mu

-0.61

cond

-0.59

zzo

-0.58

 Dunham

-0.58

POSITIVE LOGITS

erella

0.75

ahime

0.72

razil

0.71

ered

0.70

tymology

0.69

ership

0.68

ãĤ¢ãĥ«

0.66

 landfall

0.64

estones

0.64

arse

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.