INDEX

Explanations

specific locations or addresses mentioned in texts

oai_token-act-pair · gpt-3.5-turbo

locations and addresses within a text

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

OPLE

-0.72

 bacter

-0.70

 chore

-0.68

 congratulations

-0.67

 encyclopedia

-0.67

 inept

-0.67

Japan

-0.67

 nerds

-0.66

 bats

-0.66

 kitten

-0.66

POSITIVE LOGITS

 Avenue

1.48

 Boulevard

1.43

Rd

1.34

 Blvd

1.32

Ave

1.24

 Street

1.23

 Downs

1.17

 Parkway

1.14

 Drive

1.14

 Creek

1.14

Activations Density 0.231%

specific locations or addresses mentioned in texts

locations and addresses within a text

No Comments

No Known Activations