INDEX

Explanations

words related to thrilling or exciting 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

BuyableInstoreAndOnline

-0.84

sembly

-0.79

 Moder

-0.78

signed

-0.69

 Miko

-0.69

lain

-0.68

ãĥĥãĥĪ

-0.68

abee

-0.68

 unilaterally

-0.67

backer

-0.65

POSITIVE LOGITS

 thrill

3.07

Thr

2.25

 thriller

1.74

thr

1.32

 thrilling

1.24

 suspense

1.18

 grat

1.08

 adrenaline

1.06

 exhilar

1.03

rote

1.01

Activations Density 0.070%

words related to thrilling or exciting activities

No Comments

No Known Activations