INDEX

Explanations

references to something being officially recognized or accepted as valid or lawful

oai_token-act-pair · gpt-3.5-turbo

instances of the word "legitimate."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

hani

-0.79

azel

-0.73

draw

-0.73

ilan

-0.73

hire

-0.72

 Coat

-0.71

 Leaves

-0.69

ires

-0.69

hirt

-0.68

ocket

-0.67

POSITIVE LOGITS

itimate

1.06

 legitimate

0.94

soDeliveryDate

0.78

tarian

0.70

estate

0.68

ances

0.66

 sounding

0.65

DonaldTrump

0.65

 tradem

0.65

 legit

0.64

Activations Density 0.010%

references to something being officially recognized or accepted as valid or lawful

instances of the word "legitimate."

No Comments

No Known Activations