INDEX

Explanations

action verbs or words related to tasks and activities

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 referen

-1.18

acebook

-1.10

ometimes

-0.96

 predec

-0.90

avorite

-0.90

 encount

-0.88

razil

-0.84

itably

-0.83

 redes

-0.80

escription

-0.80

POSITIVE LOGITS

 Lumpur

0.96

stal

0.80

WAYS

0.76

CLASSIFIED

0.75

dain

0.74

×Ĳ

0.74

NetMessage

0.71

lust

0.71

opian

0.70

 Occupations

0.69

Activations Density 1.769%

action verbs or words related to tasks and activities

No Comments

No Known Activations