INDEX

Explanations

words related to checking or ensuring something is correct

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Ruin

-0.82

NetMessage

-0.69

ulia

-0.69

laughter

-0.68

>>>>>>>>

-0.67

gling

-0.67

 havoc

-0.66

 accuse

-0.65

 woes

-0.64

ado

-0.63

POSITIVE LOGITS

 able

1.36

 respectful

1.24

 compliant

1.24

 accessible

1.22

 aware

1.21

 comfortable

1.20

 safe

1.19

 sufficiently

1.18

 inclusive

1.17

 resilient

1.15

Activations Density 0.361%

words related to checking or ensuring something is correct

No Comments

No Known Activations