INDEX

Explanations

topics or situations where people are divided or in disagreement

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ãĥ¼ãĥĨãĤ£

-0.77

ufact

-0.60

 periphery

-0.58

 unden

-0.56

axis

-0.56

Uncommon

-0.56

wagen

-0.54

 Antar

-0.53

ongs

-0.51

Interstitial

-0.51

POSITIVE LOGITS

 about

1.33

 regarding

1.29

 whether

1.25

 ABOUT

1.18

 concerning

1.15

about

1.13

 About

1.01

how

0.97

whether

0.97

 Regarding

0.95

Activations Density 0.265%

topics or situations where people are divided or in disagreement

No Comments

No Known Activations