INDEX

Explanations

phrases related to logging into a system or account

oai_token-act-pair · gpt-3.5-turbo

terms related to login and access authentication

oai_token-act-pair · gpt-4o-mini Triggered by @bot

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ranged

-0.79

 Scha

-0.74

 Rapp

-0.70

Zam

-0.69

 paraph

-0.67

leys

-0.67

estic

-0.66

 bush

-0.66

bush

-0.65

 Noise

-0.65

POSITIVE LOGITS

 login

2.81

 Login

2.18

Login

1.87

login

1.86

 checkout

1.51

 redeem

1.35

 redemption

1.16

 redeemed

1.13

ezvous

0.93

ammu

0.88

Activations Density 0.053%

phrases related to logging into a system or account

terms related to login and access authentication

No Comments

No Known Activations