INDEX

Explanations

names related to a political figure

oai_token-act-pair · gpt-3.5-turbo

references to political events and scandals

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

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

 Shanghai

-0.95

 Czech

-0.93

Hung

-0.93

 Hung

-0.92

RI

-0.92

 Prague

-0.92

 Jiang

-0.90

 Hungarian

-0.89

Qin

-0.89

 contag

-0.89

POSITIVE LOGITS

 Moore

2.55

Moore

2.45

 Alabama

1.60

Alabama

1.44

Roy

1.26

Ala

1.23

Roy

1.21

 Luther

1.19

 Montgomery

1.19

 Leigh

1.17

Activations Density 0.182%

names related to a political figure

references to political events and scandals

No Comments

No Known Activations