INDEX

Explanations

emails or usernames

oai_token-act-pair · gpt-3.5-turbo

elements or features related to technology or digital media

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 caring

-0.69

 pinpoint

-0.67

 drunken

-0.67

 settlements

-0.66

 condensed

-0.65

 discern

-0.64

 exha

-0.64

 purposefully

-0.62

 depleted

-0.62

 deciding

-0.61

POSITIVE LOGITS

FH

1.37

kj

1.33

ZX

1.31

MQ

1.30

wx

1.28

vP

1.26

vu

1.25

UX

1.24

dq

1.23

FK

1.22

Activations Density 0.053%

emails or usernames

elements or features related to technology or digital media

No Comments

No Known Activations