INDEX

Explanations

titles or positions related to high-ranking roles in organizations

oai_token-act-pair · gpt-3.5-turbo

titles and positions indicating leadership or authority in various organizations

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

 vain

-0.83

 reper

-0.79

 shave

-0.78

 trap

-0.76

 payday

-0.76

 toget

-0.75

 nerv

-0.74

 foss

-0.74

 swarm

-0.74

 tumble

-0.73

POSITIVE LOGITS

 Manager

1.44

 Director

1.44

 Coordinator

1.42

 Officer

1.37

 Editor

1.36

 Adviser

1.32

 Secretary

1.30

 Partner

1.29

 Producer

1.29

 Minister

1.29

Activations Density 0.155%

titles or positions related to high-ranking roles in organizations

titles and positions indicating leadership or authority in various organizations

No Comments

No Known Activations