INDEX

Explanations

email subscription-related confirmation cues

oai_token-act-pair · gpt-3.5-turbo

questions and hypothetical statements

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

 engagements

-0.68

 sections

-0.67

 every

-0.64

 spoken

-0.62

 default

-0.62

 arts

-0.62

 guerrilla

-0.61

 majority

-0.61

 standing

-0.61

 hours

-0.61

POSITIVE LOGITS

 Could

3.34

Could

2.56

could

2.06

 Would

2.01

 Wouldn

1.84

 Didn

1.76

 Might

1.73

 Cannot

1.59

Would

1.57

 Possibly

1.43

Activations Density 0.014%

email subscription-related confirmation cues

questions and hypothetical statements

No Comments

No Known Activations