INDEX

Explanations

references to political offices or official positions

oai_token-act-pair · gpt-3.5-turbo

references to various "offices" associated with individuals or governmental entities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

âķĲâķĲ

-0.80

oise

-0.78

iser

-0.70

ï¸

-0.68

 Haram

-0.67

âķĲ

-0.66

MQ

-0.66

 Laksh

-0.63

ushima

-0.63

velength

-0.63

POSITIVE LOGITS

holders

1.16

holder

1.07

tops

0.93

 holder

0.88

mate

0.81

 holders

0.80

matical

0.79

room

0.78

press

0.77

 bearer

0.76

Activations Density 0.023%

references to political offices or official positions

references to various "offices" associated with individuals or governmental entities

No Comments

No Known Activations