INDEX

Explanations

capitalized words or acronyms related to military or specific geographical locations

oai_token-act-pair · gpt-3.5-turbo

instances of single-letter words or abbreviations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 Nich

-0.82

 Sally

-0.81

Wyn

-0.80

 Whites

-0.78

 Molly

-0.78

 Mania

-0.77

Nich

-0.72

Wid

-0.71

 Username

-0.71

 Nationwide

-0.71

POSITIVE LOGITS

ar

1.64

AR

1.50

ars

1.44

arp

1.28

aris

1.25

arak

1.23

ari

1.22

aro

1.20

arat

1.20

ared

1.18

Activations Density 0.149%

capitalized words or acronyms related to military or specific geographical locations

instances of single-letter words or abbreviations

No Comments

No Known Activations