INDEX

Explanations

proper names related to places and people

oai_token-act-pair · gpt-3.5-turbo

instances of specific names or locations, particularly those related to individuals or places mentioned repeatedly

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Wars

-0.83

RX

-0.81

 Decoder

-0.79

umen

-0.71

 defund

-0.70

idi

-0.70

Aid

-0.68

 Muse

-0.66

CC

-0.64

MEP

-0.64

POSITIVE LOGITS

 Hawth

4.00

 Frem

2.31

Haw

1.72

 Albion

1.69

 Carlton

1.53

 Auburn

1.41

 Hampton

1.37

 Coul

1.30

 Bellev

1.27

 Tasman

1.23

Activations Density 0.047%

proper names related to places and people

instances of specific names or locations, particularly those related to individuals or places mentioned repeatedly

No Comments

No Known Activations