INDEX

Explanations

references to electrical and power-related terms

oai_token-act-pair · gpt-3.5-turbo

references to electric companies and related terms

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

 manifesto

-0.92

 hasht

-0.78

 quota

-0.70

 avatar

-0.69

 messenger

-0.69

 surrog

-0.68

 bounty

-0.67

 martyr

-0.67

pas

-0.65

 shader

-0.65

POSITIVE LOGITS

 Electric

3.74

Electric

2.47

 Electricity

1.76

electric

1.68

 Flying

1.54

 electric

1.47

 Electro

1.46

 Electrical

1.42

 ELECT

1.38

 Nuclear

1.33

Activations Density 0.029%

references to electrical and power-related terms

references to electric companies and related terms

No Comments

No Known Activations