INDEX

Explanations

the word "called" followed by another word or phrase

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

olitics

-0.77

edia

-0.75

feat

-0.71

oday

-0.68

bilt

-0.68

iland

-0.67

isphere

-0.66

itaire

-0.64

enture

-0.64

mit

-0.64

POSITIVE LOGITS

 upon

1.17

 forth

0.93

 into

0.87

out

0.74

by

0.72

oused

0.71

 Attention

0.70

 onto

0.69

 hostage

0.69

 attention

0.69

Activations Density 0.053%

the word "called" followed by another word or phrase

No Comments

No Known Activations