INDEX

Explanations

words related to things being open to discussion or interpretation

oai_token-act-pair · gpt-3.5-turbo

words relating to debatable or questionable points and their objective qualities

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

vol

-0.66

 hemorrh

-0.66

 Franklin

-0.64

 Coast

-0.63

 winds

-0.61

Est

-0.60

 Baptist

-0.60

co

-0.60

 senior

-0.60

 rolls

-0.59

POSITIVE LOGITS

atable

4.73

isable

1.35

utable

1.28

izable

1.28

uable

1.23

ating

1.22

vable

1.21

ATED

1.21

ivable

1.20

ifiable

1.18

Activations Density 0.014%

words related to things being open to discussion or interpretation

words relating to debatable or questionable points and their objective qualities

No Comments

No Known Activations