INDEX

Explanations

names of places or people with an emphasis on specific letters or combinations

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

specific entities and names associated with individuals or groups

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

 PLUS

-0.68

 Colossus

-0.68

Pax

-0.65

 Fischer

-0.62

 Ples

-0.62

 McCl

-0.60

 exception

-0.60

 Next

-0.59

LESS

-0.58

 Rack

-0.58

POSITIVE LOGITS

hw

1.17

war

1.17

ena

1.16

than

1.13

agar

1.05

avan

1.05

amy

1.03

hi

1.02

het

1.02

heed

1.01

Activations Density 0.073%

names of places or people with an emphasis on specific letters or combinations

specific entities and names associated with individuals or groups

No Comments

No Known Activations