INDEX

Explanations

references to awards and accolades

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

phrases related to awards and accolades in the entertainment industry

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 12-res-jb

Configuration

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

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_post

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

 barracks

-0.71

STD

-0.70

 Barney

-0.68

 anus

-0.68

 bases

-0.65

 excuse

-0.65

CBI

-0.63

 reaction

-0.62

 bunker

-0.62

 centre

-0.61

POSITIVE LOGITS

winning

1.94

winner

1.54

nom

1.40

eligible

1.26

win

1.25

Win

1.22

worthy

1.21

caliber

1.21

seeking

1.15

holding

1.14

Activations Density 0.025%

references to awards and accolades

phrases related to awards and accolades in the entertainment industry

No Comments

No Known Activations